CN117676045A - Image processing parameter optimizing method and device, electronic equipment and storage medium - Google Patents
Image processing parameter optimizing method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117676045A CN117676045A CN202311370804.6A CN202311370804A CN117676045A CN 117676045 A CN117676045 A CN 117676045A CN 202311370804 A CN202311370804 A CN 202311370804A CN 117676045 A CN117676045 A CN 117676045A
- Authority
- CN
- China
- Prior art keywords
- parameter
- image processing
- optimizing
- particle
- optimal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012545 processing Methods 0.000 title claims abstract description 138
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000002245 particle Substances 0.000 claims abstract description 148
- 230000006870 function Effects 0.000 claims abstract description 47
- 230000008447 perception Effects 0.000 claims abstract description 41
- 230000016776 visual perception Effects 0.000 claims abstract description 33
- 238000012216 screening Methods 0.000 claims abstract description 15
- 238000011156 evaluation Methods 0.000 claims abstract description 14
- 238000004590 computer program Methods 0.000 claims description 5
- 238000013519 translation Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 claims 2
- 238000003786 synthesis reaction Methods 0.000 claims 2
- 239000002131 composite material Substances 0.000 claims 1
- 238000004422 calculation algorithm Methods 0.000 abstract description 37
- 238000012937 correction Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 10
- 238000012360 testing method Methods 0.000 description 10
- 239000000463 material Substances 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 230000014616 translation Effects 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000003705 background correction Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000006073 displacement reaction Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 241000023320 Luma <angiosperm> Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- OSWPMRLSEDHDFF-UHFFFAOYSA-N methyl salicylate Chemical compound COC(=O)C1=CC=CC=C1O OSWPMRLSEDHDFF-UHFFFAOYSA-N 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an image processing parameter optimizing method, an image processing parameter optimizing device, electronic equipment and a storage medium, wherein the image processing parameter optimizing method comprises the following steps: acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the optimal parameter particles as the image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters. According to the invention, the objective evaluation of the optimizing result is realized by replacing the traditional artificial subjective experience with the optimizing objective function, and the accuracy of the optimizing result is improved, so that the performance of the visual perception algorithm model is improved.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a method and apparatus for optimizing an image processing parameter, an electronic device, and a storage medium.
Background
Advanced driving assistance systems and autopilot systems for intelligent automobiles rely on vehicles to accurately perceive the surrounding environment, and thus image signal processing (Image Signal Process, ISP) technology plays a key role in autopilot or assisted driving. The environment image acquired by the vehicle-mounted camera is input into a visual perception algorithm to realize the processing of the image signal. Most of visual perception algorithms use machine vision and deep learning, and the models are black box models, so that image processing is required to be carried out on original image data before visual recognition is carried out, the image processing directly influences the output efficiency of the visual perception model, in the debugging process, an ISP engineer is required to continuously modify a large number of ISP parameters, different debugging schemes are debugged, and the debugging result depends on the visual preference of the ISP engineer. Therefore, the existing image processing parameter optimizing process has low efficiency, wastes manpower and material resources, and the accuracy of an optimizing result is difficult to objectively measure, so that the performance of a visual perception algorithm model cannot be effectively ensured.
Disclosure of Invention
The invention provides an image processing parameter optimizing method, an image processing parameter optimizing device, electronic equipment and a storage medium, which are used for solving the defects that the efficiency of the optimizing process of the existing image processing parameter optimizing method is low, the accuracy of an optimizing result is difficult to objectively measure, and the output effect of a visual perception algorithm model is difficult to guarantee.
The invention provides an image processing parameter optimizing method, which comprises the following steps:
initializing a parameter particle swarm, wherein each parameter particle in the particle swarm is an image processing parameter;
acquiring original image data, and processing the original image data by using the image processing parameters to obtain a digital image;
inputting the digital image into a visual perception model to obtain perception target information;
calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value;
updating the speed of each optimal parameter particle and the position in the parameter particle group;
and (3) taking the updated optimal parameter particles as image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters.
According to the image processing parameter optimizing method provided by the invention, the perception target information is obstacle target information in an automatic driving scene, and the optimizing target function construction method comprises the following steps:
calculating performance comprehensive scores of the plurality of performance indexes;
and constructing an optimizing objective function according to the performance comprehensive scores corresponding to different distance intervals between the obstacle target and the own vehicle.
According to the image processing parameter optimizing method provided by the invention, the optimizing objective function is the weighted summation of the performance comprehensive scores corresponding to the distance intervals, wherein the weight of each performance comprehensive score is determined according to the precision requirements corresponding to different distances.
According to the image processing parameter optimizing method provided by the invention, the performance comprehensive scores of the calculated multiple performance indexes are calculated according to the model output precision and the true positive evaluation index, and the true positive evaluation index comprises one or more of average translation error, average scale error, average direction error, average speed error and average attribute error.
According to the image processing parameter optimizing method provided by the invention, the fitness value of each parameter particle position is calculated according to the optimizing objective function, and the method comprises the following steps:
fitness value f=maxz (t), where Z (t) is the optimizing objective function.
According to the image processing parameter optimizing method provided by the invention, the updating of each optimal parameter particle speed and the position in the parameter particle swarm comprises the following steps:
taking the fitness value of each parameter particle as an individual extremum of each parameter particle, and taking the optimal fitness value of all parameter particles in each distance interval as a population extremum;
and updating the position and the speed of each parameter particle according to the individual extremum and the population extremum.
The image processing parameter optimizing method provided by the invention further comprises the following steps:
the original image data is reused to optimize the image processing parameters during each iteration.
The invention also provides an image processing parameter optimizing device, which comprises:
the initialization module is used for initializing parameter particle swarm, wherein each parameter particle in the particle swarm is an image processing parameter;
the processing module is used for acquiring original image data, and processing the original image data by using the image processing parameters to obtain a digital image;
the input module is used for inputting the digital image into a visual perception model to obtain perception target information;
the screening module is used for calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening a plurality of optimal parameter particles according to the fitness value;
the updating module is used for updating the speed of each optimal parameter particle and the position of each optimal parameter particle in the parameter particle group;
and the iteration module is used for carrying out iteration calculation on the updated optimal parameter particles as the image processing parameters again until the iteration termination condition is met, and outputting the optimal image processing parameters.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the image processing parameter optimizing method according to any one of the above when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the image processing parameter optimizing method of any one of the above.
According to the image processing parameter optimizing method, the image processing parameter optimizing device, the electronic equipment and the storage medium, each parameter particle in the particle swarm is an image processing parameter through initializing the parameter particle swarm; acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the updated optimal parameter particles as image processing parameters again for iterative computation until the iterative termination condition is met, outputting the optimal image processing parameters, realizing automatic optimization of the image processing parameters through a particle swarm algorithm, improving the parameter optimizing efficiency, saving manpower and material resources, and comparing the conventional thought subjective experience replaced by an optimizing objective function to realize objective evaluation of the optimizing result, thereby improving the accuracy of the optimizing result and improving the output effect of the visual perception algorithm model.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of an image processing parameter optimizing method provided by the invention;
FIG. 2 is a schematic diagram of an image processing parameter processing flow provided by the present invention;
FIG. 3 is a second flow chart of the image processing parameter optimizing method according to the present invention;
FIG. 4 is a third flow chart of the image processing parameter optimizing method according to the present invention;
FIG. 5 is a schematic diagram of an image processing parameter optimizing apparatus according to the present invention;
FIG. 6 is a schematic diagram of an image processing parameter optimizing system according to the present invention;
fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a flowchart of an image processing parameter optimizing method according to an embodiment of the present invention, where, as shown in fig. 1, the image processing parameter optimizing method according to the embodiment of the present invention includes:
step 101, initializing parameter particle swarms, wherein each parameter particle in the particle swarm is an image processing parameter;
102, acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image;
step 103, inputting the digital image into a visual perception model to obtain perception target information;
104, calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value;
step 105, updating the speed of each optimal parameter particle and the position in the parameter particle group;
and 106, taking the updated optimal parameter particles as image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters.
In the embodiment of the invention, the iteration termination condition is that the maximum iteration number is reached.
Traditional image processing parameter optimizing methods require ISP engineers to continuously modify a large number of ISP parameters, debug different teaching schemes, and the result of the teaching depends on human visual preference. Therefore, the efficiency of the image processing parameter optimizing process in the existing visual perception algorithm model is low, manpower and material resources are wasted, the accuracy of the optimizing result is difficult to objectively measure, and the performance of the visual perception algorithm model cannot be effectively guaranteed.
According to the image processing parameter optimizing method provided by the embodiment of the invention, the parameter particle swarm is initialized, and each parameter particle in the particle swarm is an image processing parameter; acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the updated optimal parameter particles as image processing parameters again for iterative computation until the iterative termination condition is met, outputting the optimal image processing parameters, realizing automatic optimization of the image processing parameters through a particle swarm algorithm, improving the parameter optimizing efficiency, saving manpower and material resources, and comparing the conventional thought subjective experience replaced by an optimizing objective function to realize objective evaluation of an optimizing result, thereby improving the accuracy of the optimizing result and improving the performance of a visual perception algorithm model.
Based on any of the above embodiments, the digital image is image data in YUV format, and YUV is mainly used to optimize transmission of color video signals, so that the color video signals are backwards compatible with old black-and-white televisions. The biggest advantage of this is that it takes up little bandwidth compared to RGB video signal transmission. Wherein "Y" represents brightness (Luminance or Luma), that is, a gray scale value; is a baseband signal. "U" and "V" denote Chroma (Chroma) to describe the image color and saturation for the color of the given pixel. U and V are not baseband signals, but quadrature modulated signals.
Image processing parameters in embodiments of the present invention include, but are not limited to, parameters related to the following modules: black level correction (Black Levels Ccorrect ion, BLC), noise Profile (NP), green Equalizer (GE), dynamic dead point correction (Dynamic dead pixel correct ion, dynamic DPC), auto-exposure (AE), lens shading correction (Lens shading correct ion, LSC), auto-white balance (AWB), gamma correction (Gamma), color correction matrix (Color correct ion matrix, CCM), color difference correction (Chromat ic aberration correction, CAC), denoising (de-Noise), de-mosaic (de), edge Enhancement (EE), color space conversion (colour space matrix, CSM).
As shown in fig. 2, the process of processing the original image data by the image processing parameters sequentially includes: dynamic dead pixel correction, black level correction, denoising, lens shading correction, automatic white balance, demosaicing, color correction matrix, gamma correction, edge enhancement and color space conversion, and finally obtaining the YUV format image.
In the embodiment of the invention, the original image data is repeatedly used to optimize the image processing parameters in each iteration process.
Conventionally, when the ISP engineer debugs the ISP parameters, a real vehicle test is required after each debugs, that is, the debugs scheme is downloaded to an automatic/auxiliary driving system of the vehicle, data is collected and a large number of road tests are performed to verify the performance of the perception algorithm.
In the embodiment of the invention, the existing Raw Data road test Data are repeatedly utilized, so that the need of re-acquiring Data by using manpower and material resources again each time is avoided, the cost is saved, and the efficiency is improved.
Based on any of the above embodiments, as shown in fig. 3, when the perceived target is an obstacle target in an autopilot scenario, the plurality of performance metrics include, but are not limited to, the following:
an average displacement error (Average Trans lation Error, ATE), which is the difference between the average displacement amount of the measurement object and the actual displacement amount in a certain period of time;
the average translational error (Average Trans lat ion Error, ATE) is the euclidean distance of the center of the target in the 2D plane;
the average scale error (Average Scale Error, ASE) is the 3D IoU error after alignment of the direction and translation;
the average direction error (Average Orientat ion Error, AOE) is the smallest yaw angle difference between the predicted value and the true value.
The average speed error (Average Velocity Error, AVE) is the absolute speed error of the L2 norm as a 2D (in m/s) speed difference.
Average attribute error (Average Attribute Error, AAE) is defined as 1 minus attribute classification accuracy.
In the embodiment of the invention, the construction method of the optimizing objective function comprises the following steps:
step 301, calculating performance comprehensive scores of a plurality of performance indexes:
wherein mAP is average precision, mTP is true positive evaluation index, TP is mTP set; the true positive evaluation index includes, but is not limited to, average translational error, average scale error, average direction error, average speed error, average attribute error, and the like.
And 302, constructing an optimizing objective function according to performance comprehensive scores corresponding to different distance intervals between the obstacle target and the vehicle.
In the embodiment of the invention, the optimizing objective function is the weighted summation of the performance comprehensive scores corresponding to a plurality of distance intervals, wherein the weight of each performance comprehensive score is determined according to the precision requirements corresponding to different distances.
The t=1, 2,3 … j category targets are defined as optimized obstacle targets including, but not limited to, cars, trucks, motorcycles, pedestrians, and the like.
Definition r=1, 2,3, respectively representing three distance intervals of (0, 100 m), (100 m,200 m) and (200 m,300 m), the distance referring to the distance between the own vehicle and the obstacle target.
Definition of the definitionFor the comprehensive grading of the t-th class targets in the r-distance interval, Z is defined as an optimizing target function, and because the requirements of automatic driving control on the accuracy of targets at different distances are different, the shorter the distance is, the shorter the time for the vehicle control reaction is, for example, the risk of collision possibly exists within 100m, so that the closer the distance is, the more important the image processing parameters are, and the larger the set weight value is; heddle alignment according to different distance intervalsThe score-combining weighting yields the following formula:
it should be noted that the specific value of the weight is not limited in the present invention, and those skilled in the art may choose according to actual needs.
The convergence speed of the particle swarm algorithm can be improved by weighting the comprehensive scores according to different distance intervals.
Based on any of the above embodiments, calculating the fitness value of each parameter particle position according to the optimizing objective function includes:
fitness value f=maxz (t), where Z (t) is the optimizing objective function.
Based on any of the above embodiments, updating each of the optimal parameter particle velocities and positions in the parameter particle population comprises:
taking the fitness value of each parameter particle as an individual extremum of each parameter particle, and taking the optimal fitness value of all parameter particles in each distance interval as a population extremum;
and updating the position and the speed of each parameter particle according to the individual extremum and the population extremum.
According to the embodiment of the invention, the particle swarm algorithm is utilized to optimize the objective function, and is derived from simulating the predation behavior of birds, and the simplest and effective strategy for finding the real object by the predation of birds is to search the surrounding area of the birds closest to the real object at present. The algorithm firstly initializes a group of particles in a feasible solution space, each particle represents a potential optimal solution of the extremum optimization problem, three indexes of position, speed and fitness value are used for representing the particle characteristics, the fitness value is obtained by calculation of a fitness function, and the quality of the value represents the quality of the particle. The particles move in the solution space and the individual positions are updated by tracking the individual extremum Pbest and the population extremum Gbest. The individual extremum Pbest refers to the optimal position of the fitness value calculated in the positions experienced by the individual, and the population extremum Gbest refers to the optimal position of the fitness value searched by all particles in the population. The fitness value is calculated once every time the particle is updated in position, and the individual extremum Pbest and the group extremum Gbest positions are updated by comparing the fitness value of the new particle with the fitness values of the individual extremum and the group extremum.
Assuming that D parameters need to be tuned, in a D-dimensional search space, a population x= (X) consisting of n particles 1 +X 2 ,...,X n ) Wherein the ith particle is represented as a vector X in D-dimension i ＝(x i1 +x i2 ,...，x iD ) T Representing the position of the ith particle in the D-dimensional search space and also representing one potential solution to the problem. The position X of each particle can be calculated according to the objective function i Corresponding fitness value. The speed of the ith particle is V i ＝(V i1 +V i2 ,...，V iD ) T Its individual extremum P i ＝(P i1 +P i2 ，...，P iD ) T Population extremum P of population g ＝(P g1 +P g2 ,...,P gD ) T 。
During each iteration, the particles update their own velocity and position through the individual extremum and population extremum, i.e
The speed update formula:
location update formula
The fitness function maximizes the optimizing objective function Z, i.e
F＝maxZ(t)
Wherein ω is the inertial weight, d=1, 2,. -%, D; i=1, 2,. -%, n; k is the current iteration number; v (V) id Is the velocity of the particles; c 1 And c 2 A constant that is non-negative, called an acceleration factor; r is (r) 1 And r 2 Is distributed in [0,1 ]]Random numbers of intervals. To prevent blind searching of particles, the position and speed are limited to a certain interval [ -X max ,X max ]、[-V max ,V max ]。
As shown in fig. 4, the image processing parameter optimizing method provided in the embodiment of the present invention specifically includes:
acquiring original image Data (Raw Data), carrying out two branches on the Raw Data at the same time, optimizing ISP parameters by one branch, and manually labeling by one branch by using default parameters to obtain labeling Data;
initializing particle position and velocity in optimizing ISP parameters; processing the Raw Data according to ISP parameters to obtain YUV Data; the visual perception model infers YUV data; outputting perception target information;
converging the two branches, and solving an optimizing target function according to the labeling data and the perception target information; updating individual and group optimal particles according to the solving result; and when the maximum iteration times are not reached, iteratively updating the position and the speed of each optimal particle, and when the maximum iteration times are reached, stopping iteration and outputting optimal parameters.
The image processing parameter optimizing method provided by the embodiment of the invention defines the function Z as the objective evaluation standard of the performance of the sensing algorithm and as the objective function for solving the ISP parameter optimizing algorithm, does not need ISP engineers to find the optimal parameters of the matching sensing algorithm by subjective experience, directly faces to the visual sensing algorithm of machine learning and deep learning, omits the processes of manual parameter adjustment and verification, saves a great amount of manpower and material resources, and efficiently matches the optimal ISP parameters for the sensing algorithm in the fast iteration environment of the automatic driving algorithm.
The image processing parameter optimizing device provided by the invention is described below, and the image processing parameter optimizing device described below and the image processing parameter optimizing method described above can be correspondingly referred to each other.
Fig. 5 is a schematic diagram of an image processing parameter optimizing apparatus according to an embodiment of the present invention, and as shown in fig. 5, the image processing parameter optimizing apparatus according to an embodiment of the present invention includes:
an initializing module 501, configured to initialize a parameter particle swarm, where each parameter particle in the particle swarm is an image processing parameter;
the processing module 502 is configured to obtain original image data, and process the original image data using the image processing parameters to obtain a digital image;
an input module 503, configured to input the digital image into a visual perception model to obtain perception target information;
the screening module 504 is configured to calculate an fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screen out a plurality of optimal parameter particles according to the fitness value;
an updating module 505, configured to update a speed of each optimal parameter particle and a position in the parameter particle group;
and the iteration module 506 is configured to re-use the updated optimal parameter particles as the image processing parameters to perform iterative computation until the iteration termination condition is satisfied, and output the optimal image processing parameters.
In some embodiments of the invention, further comprising
A construction module for constructing a perception target function, the construction module being configured to, when the perception target is an obstacle target in an autopilot scenario:
calculating performance comprehensive scores of the plurality of performance indexes;
and constructing an optimizing objective function according to the performance comprehensive scores corresponding to different distance intervals between the obstacle target and the own vehicle.
In an embodiment of the present invention, the processing module 502 is configured to:
in each iteration, the original image data is reused to optimize the image processing parameters.
In an embodiment of the present invention, the screening module 504 is configured to:
and calculating an fitness value F=maxZ (t) of the position of each parameter particle according to the optimizing objective function, wherein Z (t) is the optimizing objective function.
In an embodiment of the present invention, the update module 505 is configured to:
taking the fitness value of each parameter particle as an individual extremum of each parameter particle, and taking the optimal fitness value of all parameter particles in each distance interval as a population extremum;
and updating the position and the speed of each parameter particle according to the individual extremum and the population extremum.
According to the image processing parameter optimizing device provided by the embodiment of the invention, the parameter particle swarm is initialized, and each parameter particle in the particle swarm is an image processing parameter; acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the updated optimal parameter particles as image processing parameters again for iterative computation until the iterative termination condition is met, outputting the optimal image processing parameters, realizing automatic optimization of the image processing parameters through a particle swarm algorithm, improving the parameter optimizing efficiency, saving manpower and material resources, and comparing the conventional thought subjective experience replaced by an optimizing objective function to realize objective evaluation of an optimizing result, thereby improving the accuracy of the optimizing result and improving the performance of a visual perception algorithm model.
Fig. 6 is a schematic diagram of an image processing parameter optimizing system according to an embodiment of the present invention, and as shown in fig. 6, the image processing parameter optimizing system according to an embodiment of the present invention includes:
the system comprises a data acquisition system, a perception algorithm model iteration platform and an ISP parameter self-adaptive optimizing computing system;
the data acquisition system is a data acquisition system carrying an automatic driving area controller (ADU) and various sensors, and can comprehensively acquire data of a test road according to different sensor positions, wherein the data of the test road comprises high-precision map positioning information so as to train and test a visual perception model through the high-precision map positioning information. The acquired Raw Data is subjected to true value labeling and then used as a training set and a testing set of a perception algorithm, and the training set and the testing set are used for training and testing the perception algorithm model iteration platform and the ISP parameter self-adaptive optimizing computing system.
The perception algorithm model iteration platform is used for upgrading and updating the perception algorithm model, and after the model is updated, better ISP parameters need to be adapted, an iterated visual perception model is input to the ISP parameter self-adaptive optimizing computing system, and the visual perception model is computed by the ISP parameter self-adaptive optimizing computing system, and the optimal ISP parameters of the ISP parameter self-adaptive optimizing computing system are received to adapt to the new model.
The image processing parameter optimizing system provided by the embodiment of the invention defines the performance evaluation standard of the perception algorithm, and weights a plurality of perception performance indexes for the rapid convergence of the particle algorithm so as to construct a single optimizing target model, thereby realizing objective evaluation of an optimizing result, improving the accuracy of the optimizing result and improving the performance of the visual perception algorithm model.
The workflow of the image processing parameter optimizing system provided by the embodiment of the invention comprises the following steps:
the Data acquisition system is used for acquiring road Data for training a visual perception algorithm and the ISP parameter self-adaptive optimizing computing system as a training set and a testing set, the Data format is Raw Data, and the road Data are input to the visual perception algorithm model stacking platform and the ISP parameter self-adaptive optimizing computing system
The visual perception algorithm model iteration platform processes the input Raw Data through image processing parameters, identifies the processed parameters, outputs perception target information, inputs the perception target information to the ISP parameter self-adaptive optimizing computing system for ISP parameter optimizing, and computes an fitness function;
and the ISP parameter self-adaptive optimizing system receives Raw Data input by the Data acquisition system and an iterative perception model input by the visual perception model iterative platform, calculates optimal ISP parameters by utilizing an optimizing objective function and a particle swarm algorithm, and outputs an optimal ISP parameter set adapting to the round of iterative perception model.
According to the image processing parameter optimizing system provided by the embodiment of the invention, the automatic optimizing of the image processing parameters is realized through the particle swarm algorithm, the parameter optimizing efficiency is improved, manpower and material resources are saved, and compared with the conventional thought subjective experience replaced by the optimizing objective function, the objective evaluation of the optimizing result is realized, so that the accuracy of the optimizing result is improved, and the performance of a visual perception algorithm model is improved.
Fig. 7 illustrates a physical schematic diagram of an electronic device, as shown in fig. 7, which may include: processor 710, communication interface (Communications Interface) 720, memory 730, and communication bus 740, wherein processor 710, communication interface 720, memory 730 communicate with each other via communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform an image processing parameter optimization method comprising: initializing parameter particle swarms, wherein each parameter particle in the particle swarm is an image processing parameter; acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the updated optimal parameter particles as image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters.
In another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the image processing parameter optimizing method provided by the above methods, the method comprising: initializing parameter particle swarms, wherein each parameter particle in the particle swarm is an image processing parameter; acquiring original image data, and processing the original image data by using image processing parameters to obtain a digital image; inputting the digital image into a visual perception model to obtain perception target information; calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value; updating the speed of each optimal parameter particle and the position in the parameter particle group; and (3) taking the updated optimal parameter particles as image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. An image processing parameter optimizing method, comprising:
initializing a parameter particle swarm, wherein each parameter particle in the particle swarm is an image processing parameter;
acquiring original image data, and processing the original image data by using the image processing parameters to obtain a digital image;
inputting the digital image into a visual perception model to obtain perception target information;
calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening out a plurality of optimal parameter particles according to the fitness value;
updating the speed of each optimal parameter particle and the position in the parameter particle group;
and (3) taking the updated optimal parameter particles as image processing parameters again to carry out iterative computation until the iteration termination condition is met, and outputting the optimal image processing parameters.
2. The method for optimizing an image processing parameter according to claim 1, wherein the perceived target information is obstacle target information in an automatic driving scene, and the method for constructing an optimizing target function comprises:
calculating performance comprehensive scores of the plurality of performance indexes;
and constructing an optimizing objective function according to the performance comprehensive scores corresponding to different distance intervals between the obstacle target and the own vehicle.
3. The method for optimizing image processing parameters according to claim 2, wherein the optimizing objective function is a weighted sum of performance synthesis scores corresponding to a plurality of distance intervals, and the weight of each performance synthesis score is determined according to the accuracy requirements corresponding to different distances.
4. The method according to claim 2, wherein the performance composite score of the plurality of performance indexes is calculated according to the model output accuracy and a true positive evaluation index, and the true positive evaluation index includes one or more of an average translation error, an average scale error, an average direction error, an average speed error and an average attribute error.
5. The image processing parameter optimizing method according to claim 1, wherein the calculating the fitness value of each parameter particle position according to the optimizing objective function includes:
fitness value f=maxz (t), where Z (t) is the optimizing objective function.
6. The method of optimizing image processing parameters according to claim 1, wherein the updating the speed of each optimum parameter particle and the position in the parameter particle group comprises:
taking the fitness value of each parameter particle as an individual extremum of each parameter particle, and taking the optimal fitness value of all parameter particles in each distance interval as a population extremum;
and updating the position and the speed of each parameter particle according to the individual extremum and the population extremum.
7. The image processing parameter optimizing method according to claim 1, characterized by further comprising:
the original image data is reused to optimize the image processing parameters during each iteration.
8. An image processing parameter optimizing apparatus, comprising:
the initialization module is used for initializing parameter particle swarm, wherein each parameter particle in the particle swarm is an image processing parameter;
the processing module is used for acquiring original image data, and processing the original image data by using the image processing parameters to obtain a digital image;
the input module is used for inputting the digital image into a visual perception model to obtain perception target information;
the screening module is used for calculating the fitness value of each parameter particle position according to the perception target information and the optimizing target function, and screening a plurality of optimal parameter particles according to the fitness value;
the updating module is used for updating the speed of each optimal parameter particle and the position of each optimal parameter particle in the parameter particle group;
and the iteration module is used for carrying out iteration calculation on the updated optimal parameter particles as the image processing parameters again until the iteration termination condition is met, and outputting the optimal image processing parameters.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the image processing parameter optimizing method according to any one of claims 1 to 7 when executing the program.
10. A non-transitory readable storage medium having stored thereon a computer program, which when executed by a processor implements the image processing parameter optimizing method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311370804.6A CN117676045A (en) | 2023-10-20 | 2023-10-20 | Image processing parameter optimizing method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311370804.6A CN117676045A (en) | 2023-10-20 | 2023-10-20 | Image processing parameter optimizing method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117676045A true CN117676045A (en) | 2024-03-08 |
Family
ID=90079775
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311370804.6A Pending CN117676045A (en) | 2023-10-20 | 2023-10-20 | Image processing parameter optimizing method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117676045A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118333245A (en) * | 2024-06-14 | 2024-07-12 | 常州市公安局 | Public safety video monitoring point location layout method and system based on space big data |
| CN118468579A (en) * | 2024-05-21 | 2024-08-09 | 北京玄戒技术有限公司 | ISP parameter determining method and device, electronic equipment and storage medium |
-
2023
- 2023-10-20 CN CN202311370804.6A patent/CN117676045A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118468579A (en) * | 2024-05-21 | 2024-08-09 | 北京玄戒技术有限公司 | ISP parameter determining method and device, electronic equipment and storage medium |
| CN118333245A (en) * | 2024-06-14 | 2024-07-12 | 常州市公安局 | Public safety video monitoring point location layout method and system based on space big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107481292B (en) | Attitude error estimation method and device for vehicle-mounted camera | |
| US11449707B2 (en) | Method for processing automobile image data, apparatus, and readable storage medium | |
| CN117676045A (en) | Image processing parameter optimizing method and device, electronic equipment and storage medium | |
| EP3686837B1 (en) | Learning method and learning device for reducing distortion occurred in warped image generated in process of stabilizing jittered image by using gan to enhance fault tolerance and fluctuation robustness in extreme situations | |
| CN115100500A (en) | Target detection method and device and readable storage medium | |
| CN114283325B (en) | Knowledge distillation-based underwater target identification method | |
| CN110349249A (en) | Real-time dense method for reconstructing and system based on RGB-D data | |
| CN113781375A (en) | Vehicle-mounted vision enhancement method based on multi-exposure fusion | |
| CN109584299B (en) | Positioning method, positioning device, terminal and storage medium | |
| CN115063447A (en) | A video sequence-based target animal motion tracking method and related equipment | |
| CN115761549A (en) | A method and system for incremental detection and identification of weak targets with small samples of drones | |
| CN115578662A (en) | UAV front-end image processing method, system, storage medium and equipment | |
| KR102652117B1 (en) | Image processing method and image processing system | |
| CN116580369B (en) | Lane end-to-end real-time detection method for automatic driving | |
| CN117911900A (en) | Substation inspection unmanned aerial vehicle obstacle avoidance target detection method and system | |
| CN117372536A (en) | Laser radar and camera calibration method, system, equipment and storage medium | |
| CN113315995A (en) | Method and device for improving video quality, readable storage medium and electronic equipment | |
| CN110276801B (en) | Object positioning method and device and storage medium | |
| CN114529890B (en) | State detection method, device, electronic device and storage medium | |
| CN115953446A (en) | Depth estimation method, device, and electronic equipment based on error correction | |
| CN118154836A (en) | Fuel tank cover positioning model training method, fuel tank cover positioning method and device | |
| CN112396016B (en) | Face recognition system based on big data technology | |
| CN118333909B (en) | A multi-eye image acquisition and preprocessing system and method for sea surface scenes | |
| CN112233071A (en) | Multi-granularity hidden danger detection method and system based on transmission network pictures in complex environment | |
| CN110400333B (en) | Coach-type binocular stereo vision device and high-precision stereo vision image acquisition method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |