CN116524303A - Intelligent auxiliary labeling method and device for remote sensing image - Google Patents

Abstract

The application discloses an intelligent auxiliary labeling method and device for remote sensing images, and relates to the technical field of image labeling. The method comprises the following steps: constructing a total cost function according to the preprocessed remote sensing image; each pixel point of the remote sensing image is used as a node of the connected weighted graph, and the cost value of the weighted edge of the adjacent pixel points is determined according to the total cost function; the total cost function is configured to enable the cost value of the weighted edges among the pixel points on the edge of the ground object of the remote sensing image to be smaller than the preset cost value; and drawing the shortest path between adjacent mark points according to the mark points sequentially input by an operator along the ground object edge of the remote sensing image to form a preliminary contour. The method can improve the labeling efficiency of the labeling personnel, save manpower and material resources and effectively solve the problem that the labeling accuracy is reduced due to fatigue of the labeling personnel.

Description

A remote sensing image intelligent auxiliary labeling method and device

technical field

The present application relates to the technical field of image labeling, in particular to a method and device for intelligent auxiliary labeling of remote sensing images.

Background technique

Image annotation is the basis for the research and development of computer vision products. Whether it is facial recognition, drone survey, medical diagnosis or automatic driving, image annotation plays an irreplaceable role. Image annotation is the process of attaching labels to images. In the process of image annotation, one label can be applied to the entire image, or multiple labels can be applied to each group of pixels in the image. These labels are predetermined by engineers and selected to provide the computer vision model with information about the objects shown in the image. In the field of geographic information, after the remote sensing mapping image is obtained, it needs to be segmented, and various regions in the remote sensing image are segmented and labeled separately to complete the labeling. Use the marked results to provide reference for remote sensing surveying and mapping, and improve the accuracy of remote sensing surveying and mapping. In this process, a large amount of labeled data is required.

At present, the data labeling methods for image segmentation in the market are manual labeling method and intelligent labeling method. The manual labeling method requires the labeler to read and recognize the image, and then use the mouse to continuously click along the target in the picture to outline the target. The edge contour of the object, and finally correctly label the image. This traditional manual labeling method is inefficient and requires a lot of manpower, seriously affecting work efficiency. And after a long time of labeling work, the labeling accuracy rate will drop due to fatigue. Although the intelligent labeling method can overcome the shortcomings of the manual labeling method, it requires a large amount of labeling data for training before putting into work, and still needs to rely on manual labeling to provide training samples, and the quality of the training samples will directly affect the accuracy of intelligent labeling.

Contents of the invention

The embodiment of the present application provides an intelligent auxiliary labeling method for remote sensing images, which solves the problem that the manual labeling efficiency in the prior art is low and requires a lot of manpower, which seriously affects the work efficiency. Fatigue caused a decline in the accuracy of labeling. A method for assisting labelers in remote sensing image labeling was realized, which improved the work efficiency of labelers and ensured the accuracy of labeling.

In the first aspect, the embodiment of the present application provides an intelligent auxiliary labeling method for remote sensing images, including: constructing a total cost function according to the preprocessed remote sensing images; using each pixel of the remote sensing images as a node of the connected weighted graph, and Determine the cost value of the weighted edge of adjacent pixels according to the total cost function; wherein, the total cost function is configured to make the cost value of the weighted edge between pixels on the edge of the remote sensing image less than A preset cost value; according to the marking points sequentially input by the operator along the edge of the remote sensing image, the shortest path between the adjacent marking points is drawn to form a preliminary outline.

With reference to the first aspect, in a first possible implementation manner, the constructing the total cost function according to the preprocessed remote sensing image includes: obtaining the strong edge point value of the remote sensing image; calculating the pixel point gradient of the remote sensing image Amplitude and pixel gradient direction cost; performing weighted summation on the strong edge point value, the pixel gradient magnitude and the pixel gradient direction cost to form the total cost function.

With reference to the first possible implementation of the first aspect, in the second possible implementation, the acquiring the strong edge point value of the remote sensing image includes: performing edge detection on the remote sensing image; The remote sensing image is grayscaled, and the strong edge point value is determined according to the pixel value of the grayscale image of the remote sensing image; the strong edge point value formula is as follows:

In the formula, I(p) is the pixel value of the pixel point p on the edge of the object in the remote sensing image after edge detection and gray scale, Z is the strong edge point, and f _Z (p) is the value of the strong edge point.

With reference to the first aspect, in a third possible implementation manner, the drawing the shortest path between the adjacent marked points includes: iteratively calculating the first generation value between the adjacent marked points, and The mark point with the minimum value of the first generation is used as a parent node; traverse all the parent nodes on the feature edge of the remote sensing image, and connect the adjacent parent nodes to obtain the distance between the mark points shortest path.

With reference to the third possible implementation manner of the first aspect, in a fourth possible implementation manner, the sequentially calculating the first generation value between adjacent marked points includes: traversing the marked points, and Add the marked point to the node queue; use the marked point as the starting point in turn, and set the initial cost value of the starting point to 0; traverse the neighborhood nodes in the neighborhood of the starting point, and calculate the number of points added to the node queue The first-generation value from the neighborhood node to the starting point; wherein, if the neighborhood node is on the diagonal of the starting point, multiply the first-generation value by

With reference to the fourth possible implementation of the first aspect, in the fifth possible implementation, the use of the marker point with the smallest value of the first generation as the parent node includes: if the first generation value and the initial cost sum of the starting point is less than the cost value of the neighborhood node, then the first generation value is given to the neighborhood node; the starting point at this time is defined as the neighborhood node parent node, and remove the origin from the node queue.

With reference to the first aspect, in a sixth possible implementation manner, the forming the preliminary contour further includes: converting the preliminary contour into a binary image; searching for an outer contour of the preliminary contour based on the binary image, Obtaining the coordinates of the contour point set of the outer contour; converting the contour point set from pixel coordinates to geographic coordinates according to the geographic coordinate information of the remote sensing image and generating a geographic coordinate file; Correction is performed until the preliminary profile meets the accuracy requirements.

In the second aspect, the embodiment of the present application provides an intelligent auxiliary labeling device for remote sensing images, which is characterized in that it includes: a total cost function module for constructing a total cost function according to the preprocessed remote sensing images; a connected weighted graph module for Each pixel of the remote sensing image is used as a node of the connected weighted graph, and the cost value of the weighted edge of the adjacent pixel is determined according to the total cost function; wherein the total cost function is configured to make the The cost value of the weighted edge between the pixel points on the edge of the remote sensing image is less than the preset cost value; the preliminary outline module is used to draw adjacent The shortest path between the marked points forms a preliminary contour.

With reference to the second aspect, in a first possible implementation manner, the constructing the total cost function according to the preprocessed remote sensing image includes: obtaining the strong edge point value of the remote sensing image; calculating the pixel point gradient of the remote sensing image Amplitude and pixel gradient direction cost; performing weighted summation on the strong edge point value, the pixel gradient magnitude and the pixel gradient direction cost to form the total cost function.

With reference to the first possible implementation of the second aspect, in the second possible implementation, the acquiring the strong edge point value of the remote sensing image includes: performing object edge detection on the remote sensing image; The remote sensing image is grayscaled, and the strong edge point value is determined according to the pixel value of the grayscale image of the remote sensing image; the strong edge point value formula is as follows:

In the formula, I(p) is the pixel value of the pixel point p on the edge of the object in the remote sensing image after edge detection and gray scale, Z is the strong edge point, and f _Z (p) is the value of the strong edge point.

With reference to the second aspect, in a third possible implementation manner, the drawing the shortest path between the adjacent marked points includes: iteratively calculating the first generation value between the adjacent marked points, and The mark point with the minimum value of the first generation is used as a parent node; traverse all the parent nodes on the feature edge of the remote sensing image, and connect the adjacent parent nodes to obtain the distance between the mark points shortest path.

With reference to the third possible implementation manner of the second aspect, in a fourth possible implementation manner, the sequentially calculating the first-generation values between adjacent marked points includes: traversing the marked points, and Add the marked point to the node queue; use the marked point as the starting point in turn, and set the initial cost value of the starting point to 0; traverse the neighborhood nodes in the neighborhood of the starting point, and calculate the number of points added to the node queue The first-generation value from the neighborhood node to the starting point; wherein, if the neighborhood node is on the diagonal of the starting point, multiply the first-generation value by

With reference to the fourth possible implementation manner of the second aspect, in the fifth possible implementation manner, using the marker point with the smallest value of the first generation as the parent node includes: if the first generation value and the initial cost sum of the starting point is less than the cost value of the neighborhood node, then the first generation value is given to the neighborhood node; the starting point at this time is defined as the neighborhood node parent node, and remove the origin from the node queue.

With reference to the second aspect, in a sixth possible implementation manner, the forming the preliminary contour further includes: converting the preliminary contour into a binary image; searching for an outer contour of the preliminary contour based on the binary image, Obtaining the coordinates of the contour point set of the outer contour; converting the contour point set from pixel coordinates to geographic coordinates according to the geographic coordinate information of the remote sensing image and generating a geographic coordinate file; Correction is performed until the preliminary profile meets the accuracy requirements.

In a third aspect, an embodiment of the present application provides a device, the device comprising: a processor; a memory for storing processor-executable instructions; when the processor executes the executable instructions, the implementation of the first aspect Or the method described in any possible implementation manner of the first aspect.

In a fourth aspect, the embodiment of the present application provides a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium includes a computer program or instruction for storing, when the computer program or instruction is executed When, the method described in the first aspect or any possible implementation manner of the first aspect is implemented.

One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

In the embodiment of the present application, by constructing a cost function and establishing a connected weighted graph with the pixels of the remote sensing image as nodes, the edge of the feature in the remote sensing image can be obtained, and then the shortest distance between adjacent points on the edge contour of the feature can be calculated to obtain the edge of the feature preliminary outline. It effectively solves the need for annotators to focus on point-by-point annotation for a long time in the existing technology, and then realizes a method of auxiliary annotation. Annotators only need to roughly circle the edges of ground objects on remote sensing images to form a more accurate After the preliminary outline, and then manual annotation, annotated images that meet the accuracy requirements can be obtained, which can avoid the long-term point-by-point annotation by the annotator, thereby improving the efficiency of annotation and saving manpower and material resources.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the accompanying drawings that need to be used in the embodiments of the present application or in the description of the prior art. Obviously, the accompanying drawings in the following description are some of the present invention. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is a flow chart of the remote sensing image intelligent auxiliary labeling method provided by the embodiment of the present application;

Fig. 2 is the flowchart of constructing the total cost function provided by the embodiment of the present application;

FIG. 3 is a flowchart of drawing the shortest path between adjacent marking points provided by the embodiment of the present application;

Fig. 4 is a schematic diagram of an intelligent auxiliary labeling device for remote sensing images provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Apparently, the described embodiments are some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

FIG. 1 is a flowchart of a method for intelligently assisted labeling of remote sensing images provided by an embodiment of the present application, including steps 101 to 103. Among them, Fig. 1 is only an execution sequence shown in the embodiment of the present application, and does not represent the only execution sequence of the remote sensing image intelligent auxiliary labeling method. In the case where the final result can be achieved, the steps shown in Fig. 1 can be paralleled Or do it upside down.

Step 101: Construct a total cost function according to the preprocessed remote sensing image. In the embodiment of the present application, the acquired remote sensing images need to be preprocessed before constructing the total cost function. A preprocessing method is shown below, and those skilled in the art should recognize that various changes and modifications can be made to the preprocessing embodiment described here without departing from the scope and spirit of the present application. It specifically includes: filtering the acquired remote sensing image to highlight the edge of the ground object in the remote sensing image. In this step, the present application uses bilateral filtering as an example, and those skilled in the art may also use other methods to achieve the edge highlighting effect. In addition, those skilled in the art may add other denoising and smoothing operations before this step according to actual conditions.

The specific implementation of step 101 is shown in Figure 2, including steps 201 to 203, specifically:

Step 201: Obtain the strong edge point values of the remote sensing image. Specifically, edge detection of ground objects is performed on remote sensing images. In order to obtain more accurate object edges, the HED edge detection algorithm is exemplarily used here to detect the object edges of the remote sensing image. It can also be replaced by the Canny algorithm, but the accuracy of the Canny algorithm is slightly lower than that of the HED edge detection algorithm. Perform grayscale processing on the remote sensing image after edge detection, and determine the strong edge point value according to the pixel value of the grayscale image of the remote sensing image; the strong edge point value formula is as follows:

In the formula, I(p) is the pixel value of the pixel point p on the edge of the ground object in the remote sensing image after edge detection and gray scale, Z is the strong edge point, and f _Z (p) is the value of the strong edge point. The pixel value of the grayscaled remote sensing image is 0-255. In the embodiment of the present application, the edge pixel points with a pixel value above 200 are regarded as strong edge points. Strong edge point values represent edge characteristics, and pixels with higher edge characteristics have lower cost values. Therefore, the strong edge point value of the feature edge pixel point with a pixel value above 200 is 0, and the strong edge point value of the feature edge pixel point with a pixel value of 200 or below is 1. In the embodiment of the present application, Gaussian filtering may also be performed after grayscale processing to eliminate Gaussian noise in the image.

Step 202: Calculate the pixel point gradient magnitude and pixel point gradient direction cost of the remote sensing image. In the embodiment of this application, the pixel point gradient magnitude f _G (p), the formula is as follows:

In the formula, f _G (p) is the magnitude of the gradient of the pixel point, G is the gradient of the pixel point, p is the pixel point, sx, sy are the gradients of the remote sensing image in the x direction and y direction, respectively. The larger the pixel gradient, the more obvious the edge features, and the lower the cost value.

Pixel gradient direction cost f _D (p,q), the formula is as follows:

In the formula, f _D (p, q) is the gradient direction cost of the pixel point, p and q are the pixel point, D is the gradient direction of the pixel point, dp is the product of the direction gradient of the pixel point p and the vector of the pixel point p and q, dq is the product of the directional gradient of pixel q and the vectors of pixel p and q. The pixel gradient direction cost can play a role in edge smoothing, so set a higher cost value at the place where the boundary changes sharply, and set a lower cost value at the place where the boundary change is gentle.

Step 203: Perform weighted summation of the strong edge point value, pixel point gradient magnitude and pixel point gradient direction cost to form a total cost function. In the embodiment of this application, the total cost function is as follows:

L _(p,q) =w ₁ f _Z (p)+w ₂ f _G (p)+w ₃ f _D (p,q)

In the formula, f _Z (p) is the strong edge point value, f _G (p) is the pixel gradient magnitude, f _D (p,q) is the pixel gradient direction cost, p and q are two pixel points, w ₁ , w ₂ , w ₃ are weights. In the embodiment of the present application, w ₁ +w ₂ +w ₃ =1, and 0.7≤w ₁ ≤0.83, 0.03≤w ₂ ≤0.05, 0.12≤w ₃ ≤0.25; for example, w ₁ =0.83, w ₂ = 0.03, w ₃ = 0.14 can achieve the better effect of this application.

Step 102: Take each pixel of the remote sensing image as a node of the connected weighted graph, and determine the cost value of the weighted edge of the adjacent pixel according to the total cost function. Wherein, the total cost function is configured so that the cost value of the weighted edge between the pixel points on the edge of the remote sensing image is smaller than the preset cost value. Specifically, each pixel of the remote sensing image is used as a node to construct a connected weighted graph, and each node of the connected weighted graph is assigned a different cost value. There is a weighted edge between adjacent pixels, and the cost value of the weighted edge Computed from the defined total cost function.

In addition, the purpose of image labeling is to find the edge contour of the ground object and then generate a label for training the neural network model, so the cost function of the shortest path is closely related to the edge features, and the cost value between pixels with more prominent edge features is lower , to ensure that the shortest path is as close as possible to the edge of the object. In order to ensure that the edge of the ground object in the remote sensing image is as close as possible to the ground object, when defining the cost function, it is necessary to ensure that the cost value of the weighted edge between the pixels on the edge of the ground object is less than the preset cost value. Here, the preset The cost value is set by those skilled in the art based on experience or multiple experimental results.

Step 103: Draw the shortest path between adjacent marked points according to the marked points sequentially input by the operator along the edge of the remote sensing image to form a preliminary outline. In the embodiment of the present application, the mouse is moved along the edge of the feature, and the position of the mouse is the marked point, and the shortest path between adjacent marked points is drawn. The specific steps of drawing the shortest path between adjacent marked points are as shown in Figure 3, including step 301 to step 302, as follows:

Step 301: iteratively calculate the first-generation value between adjacent marker points, and use the marker point with the smallest first-generation value as a parent node. Iteratively computes the first-generation values between adjacent marker points. Specifically, traverse the marked points, and add the marked points to the node queue; take the marked points as the starting point in turn, and set the initial cost value of the starting point to 0; traverse the neighborhood nodes in the neighborhood of the starting point, and calculate the neighborhood that is added to the node queue The first-generation value from the node to the starting point; among them, if the neighbor node is on the diagonal of the starting point, the first-generation value is multiplied by In the embodiment of the present application, the neighborhood nodes are eight pixel points in the eight directions of up, down, left, right, and four diagonal directions around the starting point neighborhood.

Take the first-generation marker with the smallest value as the parent node. Specifically, if the sum of the first generation value and the initial cost of the starting point is less than the cost value of the neighborhood node, the first generation value is given to the neighborhood node; the starting point at this time is defined as the parent node of the neighborhood node, and The origin is removed from the node queue. In the embodiment of this application, the sum of the first generation value and the initial cost of the starting point is compared with the cost value of the neighborhood nodes in the neighborhood. Since the initial cost value of the starting point is set to 0, here is actually the first generation value and The cost value of the neighborhood node is compared, if the first generation value is less than the initial cost value of the node in the domain, the first generation value is given to the domain node, the starting point at this time is defined as the parent node of the neighborhood node, and the starting point Dequeue a node. Process all markers in turn until all markers are iteratively calculated, that is, the node queue in this step is empty.

Step 302: traverse all parent nodes on the edge of the remote sensing image, and connect adjacent parent nodes to obtain the shortest path between the marked points. In the embodiment of the present application, the mouse is used to backtrack along the edge of the feature, and the parent node of the current edge node is constantly searched for. Until all parent nodes are found, connect adjacent parent nodes to get the shortest path between adjacent marked points. The connection edges between all adjacent parent nodes constitute the preliminary outline of the ground object.

In the embodiment of the present application, after obtaining the preliminary outline of the ground object, the preliminary outline is converted into a binary image; based on the binary image, the outer outline of the initial outline is searched to obtain the outline point set coordinates of the outer outline; according to the remote sensing image The geographic coordinate information converts the contour point set from pixel coordinates to geographic coordinates and generates a geographic coordinate file; according to the corresponding manual operation, the geographic coordinate file is revised until the preliminary outline meets the accuracy requirements. Exemplarily, the geographical coordinate file may be an SHP file, a JSON file, a GML file, and the like. The formed preliminary outline is already relatively accurate. At this time, it is only necessary to manually repair and annotate the preliminary outline, and the adjusted part can obtain an annotation file that meets the standard degree requirements.

Although the present application provides method operation steps such as embodiments or flowcharts, more or less operation steps may be included based on routine or non-inventive efforts. The order of steps listed in this embodiment is only one way of execution order of many steps, and does not represent the only execution order. When executed by an actual device or client product, the methods shown in this embodiment or the drawings may be executed sequentially or in parallel (for example, in a parallel processor or multi-thread processing environment).

As shown in FIG. 4 , the embodiment of the present application also provides an apparatus 400 for intelligent auxiliary labeling of remote sensing images. The device includes: a total cost function module 401 , a connectivity weighted graph module 402 and a preliminary contour module 403 .

The total cost function module 401 is used to construct a total cost function according to the preprocessed remote sensing image. The total cost function module 401 is specifically used to: obtain the strong edge point value of the remote sensing image; calculate the pixel gradient magnitude and pixel gradient direction cost of the remote sensing image; calculate the strong edge point value, pixel gradient magnitude and pixel gradient direction The costs are weighted and summed to form the total cost function.

The connected weighted graph module 402 is used to use each pixel point of the remote sensing image as a node of the connected weighted graph, and determine the cost value of the weighted edge of adjacent pixel points according to the total cost function. Wherein, the total cost function is configured so that the cost value of the weighted edge between the pixel points on the edge of the remote sensing image is smaller than the preset cost value. The connected weighted graph module 402 is specifically used for:

The preliminary outline module 403 is used to draw the shortest path between adjacent marked points according to the marked points sequentially input by the operator along the edge of the remote sensing image to form a preliminary outline. The preliminary outline module 403 is specifically used to: iteratively calculate the first-generation value between adjacent marker points, and use the marker point with the smallest first-generation value as the parent node; traverse all parent nodes on the feature edge of the remote sensing image, and Connect adjacent parent nodes to get the shortest path between marked points. Iteratively calculate the first-generation value between adjacent marker points, specifically including: traversing the marker points and adding the marker points to the node queue; taking the marker points as the starting point in turn, and setting the initial cost value of the starting point to 0; traversing the starting point Neighborhood nodes in the neighborhood, calculate the first-generation value from the neighborhood node that joins the node queue to the starting point; among them, if the neighborhood node is on the diagonal of the starting point, multiply the first-generation value by Take the mark point with the smallest value of the first generation as the parent node, specifically including: if the sum of the value of the first generation and the initial cost of the starting point is less than the cost value of the neighborhood node, assign the value of the first generation to the neighborhood node; The start point of is defined as the parent node of the neighborhood node, and the start point is removed from the node queue. Forming the preliminary contour also includes: converting the preliminary contour into a binary image; searching the outer contour of the preliminary contour based on the binary image to obtain the coordinates of the contour point set of the outer contour; The pixel coordinates are converted into geographic coordinates and a geographic coordinate file is generated; the geographic coordinate file is revised and annotated according to the corresponding manual operation until the preliminary outline meets the accuracy requirements.

Some of the modules of the apparatus described herein may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

The devices or modules described in the embodiments of the above application may be specifically implemented by computer chips or entities, or by products with certain functions. For the convenience of description, when describing the above devices, the functions are divided into various modules and described separately. When implementing the embodiments of the present application, the functions of each module may be implemented in one or more pieces of software and/or hardware. Of course, a module that realizes a certain function may also be implemented by a combination of multiple submodules or subunits.

The methods, devices or modules described in this application can be implemented in computer readable program code. The controller can be implemented in any suitable way. Computer-readable media for computer-readable program code (such as software or firmware) executed by a device, logic gates, switches, application-specific integrated circuits (English: Application Specific Integrated Circuit; abbreviated: ASIC), programmable logic controllers, and embedded microcontrollers Examples of the controller include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as a part of the control logic of the memory. Those skilled in the art also know that, in addition to realizing the controller in a purely computer-readable program code mode, it is entirely possible to make the controller use logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded The same function can be realized in the form of a microcontroller or the like. Therefore, this kind of controller can be regarded as a hardware component, and the devices included in it for realizing various functions can also be regarded as the structure in the hardware component. Or even, means for realizing various functions can be regarded as a structure within both a software module realizing a method and a hardware component.

The embodiment of the present application also provides a device, the device includes: a processor; a memory for storing processor-executable instructions; when the processor executes the executable instructions, the implementation as described in the embodiment of the present application Methods.

The embodiment of the present application also provides a non-volatile computer-readable storage medium, on which a computer program or instruction is stored, and when the computer program or instruction is executed, the method as described in the embodiment of the present application is executed accomplish.

In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, each module may exist independently, or two or more modules may be integrated into one module.

The above-mentioned storage medium includes but not limited to random access memory (English: Random Access Memory; abbreviation: RAM), read-only memory (English: Read-Only Memory; abbreviation: ROM), cache (English: Cache), hard disk (English: Hard Disk Drive (abbreviation: HDD) or memory card (English: Memory Card). The memory may be used to store computer program instructions.

It can be known from the above description of the implementation manners that those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary hardware. Based on this understanding, the essence of the technical solution of this application or the part that contributes to the existing technology can be embodied in the form of software products, or it can be reflected in the implementation process of data migration. The computer software product can be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions to make a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) execute this Apply the methods described in each embodiment or some parts of the embodiments.

The various implementations in this specification are described in a progressive manner, the same or similar parts of the various implementations can be referred to each other, and each implementation focuses on the differences from other implementations. This application, in whole or in part, can be used in numerous general purpose or special purpose computer system environments or configurations. Examples: personal computers, server computers, handheld or portable devices, tablet-type devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, etc.

The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be described in the foregoing embodiments Modifications to the technical solutions, or equivalent replacement of some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the present application.

Claims (10)

1. A remote sensing image intelligent auxiliary labeling method, characterized in that it comprises: Construct the total cost function according to the preprocessed remote sensing image; Using each pixel of the remote sensing image as a node of the connected weighted graph, and determining the cost value of the weighted edge of the adjacent pixel according to the total cost function; Wherein, the total cost function is configured to make the cost value of the weighted edge between the pixel points on the edge of the remote sensing image less than the preset cost value; According to the marking points sequentially input by the operator along the edge of the remote sensing image, the shortest path between the adjacent marking points is drawn to form a preliminary outline. 2. The method according to claim 1, wherein said constructing a total cost function according to the preprocessed remote sensing image comprises: Obtaining the strong edge point value of the remote sensing image; Calculate the pixel point gradient magnitude and pixel point gradient direction cost of the remote sensing image; A weighted summation is performed on the strong edge point value, the pixel point gradient magnitude and the pixel point gradient direction cost to form the total cost function. 3. The method according to claim 2, wherein said obtaining the strong edge point value of said remote sensing image comprises: performing feature edge detection on the remote sensing image; Carry out grayscale processing to described remote sensing image, and determine strong edge point value according to the pixel value of the gray scale image of described remote sensing image; Described strong edge point value formula is as follows: In the formula, I(p) is the pixel value of the pixel point p on the edge of the object in the remote sensing image after edge detection and gray scale, Z is the strong edge point, and f _Z (p) is the value of the strong edge point. 4. The method according to claim 1, wherein said drawing the shortest path between adjacent said marked points comprises: Iteratively calculating the first-generation value between adjacent marked points, and using the marked point with the smallest first-generation value as a parent node; Traverse all the parent nodes on the edge of the feature in the remote sensing image, and connect the adjacent parent nodes to obtain the shortest path between the marked points. 5. The method according to claim 4, wherein said sequentially calculating the first generation value between adjacent said marking points comprises: Traversing the marked points, and adding the marked points to the node queue; Taking the marked points as the starting point in turn, and setting the initial cost value of the starting point to 0; Traversing the neighborhood nodes in the neighborhood of the starting point, calculating the first generation value of the neighborhood node added to the node queue to the starting point; wherein, if the neighborhood node is on the diagonal of the starting point , then multiply the first-generation value by 6. The method according to claim 5, wherein said using the mark point with the smallest value of the first generation as a parent node comprises: If the sum of the first generation value and the initial cost of the starting point is less than the cost value of the neighborhood node, assign the first generation value to the neighborhood node; The starting point at this time is defined as the parent node of the neighborhood node, and the starting point is removed from the node queue. 7. The method according to claim 1, wherein said forming a preliminary profile further comprises: converting said preliminary profile into a binary image; searching the outer contour of the preliminary contour based on the binary image, and obtaining the coordinates of the contour point set of the outer contour; converting the contour point set from pixel coordinates to geographic coordinates according to the geographic coordinate information of the remote sensing image and generating a geographic coordinate file; According to the corresponding manual operation, the geographic coordinate file is corrected until the preliminary outline meets the accuracy requirement. 8. An intelligent auxiliary labeling device for remote sensing images, characterized in that it comprises: The total cost function module is used to construct the total cost function according to the preprocessed remote sensing image; The connected weighted graph module is used to use each pixel of the remote sensing image as a node of the connected weighted graph, and determine the cost value of the weighted edge of the adjacent pixel according to the total cost function; wherein, the total cost function It is configured to make the cost value of the weighted edge between the pixel points on the edge of the remote sensing image less than the preset cost value; The preliminary outline module is used to draw the shortest path between adjacent marked points according to the marked points sequentially input by the operator along the edge of the remote sensing image to form a preliminary outline. 9. A device for performing an intelligent auxiliary labeling method for remote sensing images, characterized in that it comprises: processor; memory for storing processor-executable instructions; When the processor executes the executable instructions, the method according to any one of claims 1-7 is implemented. 10. A non-volatile computer-readable storage medium, characterized in that it includes a computer program or instruction for storing, and when the computer program or instruction is executed, the computer program or instruction according to any one of claims 1 to 7 method is implemented.