TWI846177B - Road image segmentation method, computer device and storage medium - Google Patents

Abstract

The present application relates to an image analysis technology, and provides a road image segmentation method, a computer device and a storage medium. This method includes: acquiring a road image, obtaining a plurality of regions of interest by processing the road image; generating a splicing region based on the plurality of regions of interest, and obtaining a road segmentation image and a segmentation result based on the road segmentation image by inputting the splicing region to a pre-trained road segmentation model. By utilizing the present application, the segmentation accuracy of the roads in the image can be improved.

Description

Road segmentation method, computer equipment and storage medium

The present invention relates to the field of image processing, and in particular to a road segmentation method, a computer device and a storage medium.

In the current image-based road detection solution, the road in the image is blurred due to the high brightness of the road image taken in a backlit environment, resulting in the inability to accurately detect the road from the backlit road image, thus affecting driving safety.

In view of the above, it is necessary to provide a road segmentation method, computer equipment and storage medium to solve the technical problem of being unable to accurately detect the road from the backlit road image, thereby improving driving safety.

This application provides a road segmentation method, which includes: obtaining a road image, obtaining multiple regions of interest based on the road image, generating a spliced region based on the multiple regions of interest, inputting the spliced region into a pre-trained road segmentation model, and obtaining a road segmentation image and a segmentation result in the road segmentation image.

According to an optional embodiment of the present application, the multiple regions of interest include a preprocessing region, a first equalization region, a second equalization region, and a target region. The step of obtaining the multiple regions of interest based on the road image includes: preprocessing the road image to obtain the preprocessing region, performing equalization processing on the preprocessing region to obtain the first equalization region and the second equalization region, detecting the first equalization region to obtain a target edge line, and generating the target region based on the target edge line and the first equalization region.

According to an optional embodiment of the present application, the detecting the first equalized region to obtain the target edge line includes: performing edge detection on the first equalized region to obtain multiple initial edge lines, calculating the polar coordinate equation of each pixel point on each initial edge line based on the pixel position in the first equalized region, and selecting the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations.

According to an optional embodiment of the present application, the edge detection of the first equalized area to obtain multiple initial edge lines includes: filtering the first equalized area to obtain a filter area, calculating the gradient value of each pixel in the filter area, filtering multiple pixels corresponding to the multiple gradient values according to a preset upper limit threshold value and a preset lower limit threshold value to obtain edge pixels, and determining multiple lines formed by multiple adjacent edge pixels as the multiple initial edge lines.

According to an optional embodiment of the present application, the target edge line includes a first target edge line and a second target edge line, and the selecting the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations includes: drawing a curve corresponding to each polar coordinate equation, selecting the polar angle of each initial edge line based on the multiple curves, calculating the edge line slope of the corresponding initial edge line according to the polar angle, selecting the initial edge line corresponding to the maximum edge line slope as the first target edge line, and selecting the initial edge line corresponding to the minimum edge line slope as the second target edge line.

According to an optional embodiment of the present application, generating the target area according to the target edge line and the first equalization area includes: determining the area enclosed by the first target edge line and the second target edge line in the first equalization area as a road area, and binarizing the first equalization area based on the road area to obtain the target area.

According to an optional embodiment of the present application, generating a stitching region based on the multiple regions of interest includes: multiplying the pixel values of the first equalized region and the corresponding pixel points in the target region to obtain a first pixel value, and adjusting the pixel value of each pixel point in the first equalized region to the corresponding first pixel value to obtain a first region, and multiplying the pixel values of the second equalized region and the corresponding pixel points in the target region to obtain a second region. Pixel value, and adjust the pixel value of each pixel in the second equalization area to the corresponding second pixel value to obtain the second area, multiply the pixel values of the corresponding pixels in the pre-processing area and the target area to obtain a third pixel value, and adjust the pixel value of each pixel in the pre-processing area to the corresponding third pixel value to obtain the third area, and splice the first area, the second area and the third area to obtain the spliced area.

According to an optional embodiment of the present application, before inputting the spliced area into a pre-trained road segmentation model, the method further includes: obtaining a road segmentation network, a road training image, and an annotation result of the road training image, inputting the road training image into the road segmentation network for feature extraction to obtain a road feature map, performing road prediction on each pixel in the road feature map to obtain a prediction result of the road feature map, and adjusting the parameters of the road segmentation network according to the prediction result and the annotation result to obtain a trained road segmentation model.

This application provides a computer device, which includes: a memory storing at least one instruction; and a processor executing the at least one instruction to implement the road segmentation method.

This application provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in a computer device to implement the road segmentation method.

It can be seen from the above technical solution that the present application performs image processing on the road image to obtain multiple regions of interest. The image processing includes image cropping, binarization and equalization, road detection and Hough transform. Since image cropping can delete the part of the road image that does not contain the road, it can reduce the interference to the road detection. When the road image is a backlit image, by binarizing and equalizing the backlit image, the brightness of the backlit image can be reduced to make the backlit image clearer. By performing road detection on the backlit image, multiple possible initial edge lines of the road in the backlit image can be detected, and the multiple possible edge lines are screened by Hough transform to obtain the target edge line, thereby improving the accuracy of road detection. In addition, since more road information is integrated into the multiple regions of interest, the roads in the spliced regions can be made clearer by generating a spliced region based on the multiple regions of interest. In addition, the roads in the spliced region are detected by using a pre-trained road segmentation model. Since the road segmentation model has learned the location features of the roads, the location of the roads in the spliced region can be predicted more accurately.

1: Computer equipment

12: Storage

13: Processor

S11~S14, S121~S124, S141~S144: Steps

Figure 1 is an application scenario diagram of a road segmentation method provided by the embodiment of this application.

Figure 2 is a flow chart of a road segmentation method provided by the embodiment of this application.

Figure 3 is a flowchart of image processing provided by the embodiment of this application.

Figure 4 is a schematic diagram of the initial edge line and the target edge line provided in the embodiment of this application.

Figure 5 is a training flowchart of the road segmentation model provided in the embodiment of this application.

Figure 6 is a schematic diagram of a road segmentation image provided by the embodiment of this application.

Figure 7 is a schematic diagram of the structure of the computer device provided in the embodiment of this application.

In order to make the purpose, technical solution and advantages of this application clearer, this application is described in detail below with reference to the attached drawings and specific embodiments.

As shown in FIG1 , it is an application scene diagram of a road segmentation method provided by an embodiment of the present application. The road scene diagram can be obtained by shooting a road scene against light with a shooting device. Since it is against light, the image brightness of the road scene diagram is very large, resulting in blurry roads in the diagram. The application scene diagram includes roads, vehicles, trees, and the sky, etc. The road segmentation method can be applied to one or more computer devices 1, such as the computer device 1 shown in FIG7 . The computer device 1 communicates with the shooting device, which can be a monocular camera or other device with shooting function.

The computer device 1 is a device that can automatically perform parameter value calculation and/or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to: microprocessor, application specific integrated circuit (ASIC), field-programmable gate array (FPGA), digital signal processor (DSP), embedded device, etc. The computer device 1 can be any computer product that can interact with the user, such as a personal computer, tablet computer, smart phone, personal digital assistant (PDA), game console, interactive model television (IPTV), wearable smart device, etc.

The computer device 1 may also include a model device and/or a user device. The model device includes, but is not limited to, a single model server, a server group consisting of multiple model servers, or a cloud consisting of a large number of hosts or model servers based on cloud computing. The computer device 1 may also be a vehicle-mounted device in a vehicle. The model in which the computer device 1 is located includes, but is not limited to: the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

The computer device 1 may also be a vehicle-mounted device. When the computer device 1 is a vehicle-mounted device, the shooting device may be a vehicle-mounted shooting device in the vehicle-mounted device, such as a vehicle-mounted camera or a driving recorder, and the vehicle-mounted device communicates with the vehicle-mounted shooting device.

As shown in FIG2, FIG2 is a flow chart of a road segmentation method provided in an embodiment of the present application. According to different requirements, the order of each step in the flow chart can be adjusted according to the actual detection requirements, and some steps can be omitted. The execution subject of the method is a computer device.

Step S11, obtaining road images.

In at least one embodiment of the present application, the road image is a three-primary color light (Red Green Blue, RGB) image, and the road image may include objects such as vehicles, roads, lane lines, pedestrians, sky, trees, etc.

In at least one embodiment of the present application, the computer device acquires the road image including: the computer device controls the shooting device to shoot the road scene to obtain the road image. The shooting device may be a monocular camera or a vehicle-mounted camera, etc., and the road scene may include objects such as vehicles, roads, lane lines, pedestrians, sky, trees, etc.

In this embodiment, the road image includes a backlit road image, and the backlit road image is an image obtained by the shooting device shooting the road scene against the light.

Step S12, based on the road image, obtain multiple areas of interest.

In at least one embodiment of the present application, multiple regions of interest are obtained through one or more preset image processing methods, wherein the image processing methods include cropping, road detection, Hough transform, histogram equalization, and binarization processing, etc. The histogram equalization processing includes grayscale histogram equalization processing and three-channel histogram equalization processing, etc.

Specifically, the image processing based on the road image can refer to the image processing flow chart of the embodiment shown in FIG3.

Step S121, pre-process the road image to obtain the pre-processed area.

In this embodiment, the road image is preliminarily detected to determine the position of the road in the road image. For example, in the road image, the road is mainly located in the lower half of the road image, and the upper half of the road portion is mainly the distant view and the sky. At this time, the computer device can select the lower half of the road image as the pre-processing area. For example, the road image can be preliminarily detected by a target detection algorithm. The target detection algorithm can be R-CNN, Fast R-CNN, Faster R-CNN, etc.

In this embodiment, since image cropping can delete the portion of the road image that does not contain the road, interference with road detection can be reduced.

Step S122, performing equalization processing on the pre-processed area to obtain the first equalized area and the second equalized area.

In this embodiment, the computer device performs grayscale histogram equalization processing on the preprocessed area to obtain the first equalized area, and then the computer device performs histogram equalization processing on each channel of the preprocessed area to obtain the second equalized area. In this embodiment, when the road image is a backlit road image, grayscale histogram equalization processing and three-channel histogram equalization processing are performed on the preprocessed area generated after clipping, which can reduce the brightness of the first equalized area and the second equalized area, thereby reducing the impact of brightness on road detection.

Step S123, detect the first equalization area to obtain the target edge line.

In this embodiment, the computer device detects the first equalization area to obtain the target edge line, including: the computer device performs road detection on the first equalization area to obtain multiple initial edge lines, and then the computer device calculates the polar coordinate equation of each pixel point on each initial edge line based on the pixel position in the first equalization area, and further, the computer device selects the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations. The target edge line includes a first target edge line and a second target edge line, and the pixel position includes a horizontal coordinate value and a vertical coordinate value. The format of the polar coordinate equation is ρ = xcosθ + ysinθ , ρ represents the dependent variable polar diameter, θ represents the independent variable polar angle, x represents the horizontal coordinate value of each pixel point, and y represents the vertical coordinate value of each pixel point.

Specifically, the computer device performs road detection on the first equalization area to obtain multiple initial edge lines, including: the computer device performs filtering processing on the first equalization area to obtain a filtering area; the computer device calculates the gradient value of each pixel in the filtering area; the computer device filters multiple pixels corresponding to the multiple gradient values according to a preset upper limit valve value and a preset lower limit valve value to obtain edge pixels. Then, the computer device determines multiple lines formed by multiple adjacent edge pixels as the multiple initial edge lines.

Wherein, the computer device can calculate the gradient value through the Sobel operator (Sobel). The upper limit valve value and the lower limit valve value can be set by yourself, and this application does not impose any restrictions on this. For example, the upper limit valve value can be 80, and the lower limit valve value can be 40.

In this embodiment, by performing road detection on the first equalization area, the initial edge lines of multiple possible roads in the first equalization area can be preliminarily detected.

Specifically, the computer device selects a plurality of corresponding convolution pixels from the first equalization region according to a preset convolution kernel size, and generates a corresponding weight according to the pixel position of each convolution pixel. The computer device normalizes the plurality of weights to generate a Gaussian filter kernel, and then the computer device uses the Gaussian filter kernel to perform convolution on the first equalization region to obtain the filter region.

其中，h(x,y)表示每個卷積像素點對應的權重，x表示每個卷積像素點的橫座標值，y表示每個卷積像素點的縱座標值，σ為預設的標準差。所述卷積核尺寸以及所述標準差可以自行設置，本申請實施例對此不作限制。例如，所述卷積核尺寸可以為3*3，所述標準差可以為1。 The calculation formula for each weight is:

Wherein, h ( x, y ) represents the weight corresponding to each convolution pixel, x represents the horizontal coordinate value of each convolution pixel, y represents the vertical coordinate value of each convolution pixel, and σ is the preset standard deviation. The convolution kernel size and the standard deviation can be set by oneself, and the embodiment of the present application does not limit this. For example, the convolution kernel size can be 3*3, and the standard deviation can be 1.

In one embodiment, by filtering the first equalization area, the noise in the first equalization area can be filtered out, so that the edge line of the road in the filtering area is clearer.

Specifically, the computer device screens multiple pixel points corresponding to the multiple gradient values according to a preset upper limit valve value and a preset lower limit valve value, and obtains edge pixel points, including: the computer device determines the pixel points corresponding to the gradient value greater than or equal to the upper limit valve value as the edge pixel points, and determines the pixel points corresponding to the gradient value between the upper limit valve value and the lower limit valve value as the middle pixel points. If the middle pixel point is directly connected to any of the edge pixel points, the computer device determines the middle pixel point as the edge pixel point, or, if the middle pixel point is not connected to any edge pixel point, the computer device discards the middle pixel point. In addition, the computer device discards the pixel points corresponding to the gradient value less than or equal to the lower limit valve value.

In this embodiment, since the pixel values of edge pixels differ greatly from the pixel values of other pixels in the first equalization area, and the gradient value can characterize the magnitude of the change in pixel value, the accuracy of confirming edge pixels can be improved by using the pixel corresponding to the gradient value greater than the upper limit value as the edge pixel.

Specifically, the computer device selects the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations, including: the computer device draws a curve corresponding to each polar coordinate equation, and selects the polar angle of each initial edge line based on the multiple curves, the computer device calculates the edge line slope of the corresponding initial edge line according to the polar angle, and selects the initial edge line corresponding to the largest edge line slope as the first target edge line, and selects the initial edge line corresponding to the smallest edge line slope as the second target edge line.

Among them, since the dependent variable in each polar coordinate equation is the polar diameter and the argument is the polar angle, therefore, the coordinates of each point on the curve are the polar diameter and the polar angle, and the computer device selects the polar angle corresponding to the intersection of the multiple curves as the polar angle corresponding to each initial edge line. In this embodiment, the sign of the edge line slope has positive and negative signs. Therefore, the sign of the edge line slope of the first target edge line is positive, and the sign of the edge line slope of the second target edge line is negative.

In this embodiment, by selecting the target edge line from the multiple initial edge lines, the road area in the first equalization area can be preliminarily determined.

As shown in FIG. 4 , it is a schematic diagram of the initial edge line and the initial edge line provided by the embodiment of the present application. There are multiple initial edge lines in FIG. 4 (a). The slope of each initial edge line is different. Since the edge line slope of the leftmost initial edge line in FIG. 4 (a) is the largest, and the edge line slope of the rightmost initial edge line is the smallest, therefore, the leftmost initial edge line in FIG. 4 (a) is determined as the first target edge line, and the rightmost initial edge line is determined as the second target edge line, and the target edge line in FIG. 4 (b) is obtained.

In other embodiments of the present application, other edge detection algorithms may be used to perform road detection on the first equalization area. For example, the Canny edge detection algorithm.

Step S124, generating the target area according to the target edge line and the first equalization area.

In this embodiment, the computer device generates the target area according to the target edge line and the first equalization area, including: the computer device determines the area enclosed by the first target edge line and the second target edge line in the first equalization area as a road area, and performs binarization processing on the first equalization area based on the road area to obtain the target area.

For example, the computer device may adjust the pixel values of all pixels in the road area to 1, and adjust the pixel values of pixels in the first equalization area except the road area to 0.

In this embodiment, since the first target edge line is the initial edge line corresponding to the maximum edge line slope and the second target edge line is the initial edge line corresponding to the minimum edge line slope, it is possible to ensure that all road areas in the first equalized area are selected. In addition, by binarizing the first equalized area, the road area and the non-road area in the first equalized area are distinguished, making the road area in the target area clearer.

In at least one embodiment of the present application, the multiple regions of interest include a preprocessing region, a first equalization region, a second equalization region, and a target region. The preprocessing region refers to a region generated after clipping the road image, the first equalization region refers to a region generated after grayscale histogram equalization is performed on the preprocessing region, the second equalization region refers to a region generated after histogram equalization is performed on each channel of the preprocessing region, and the target region refers to a region generated after screening the initial edge line in the first equalization region.

Step S13, generating a splicing region based on multiple regions of interest.

In at least one embodiment of the present application, the computer device generates a stitching area based on the multiple regions of interest, including: the computer device multiplies the pixel values of the corresponding pixels in the first equalized area and the target area to obtain a first pixel value, and adjusts the pixel value of each pixel in the first equalized area to the corresponding first pixel value to obtain a first area; the computer device multiplies the pixel values of the corresponding pixels in the second equalized area and the target area to obtain The computer device calculates the second pixel value, and adjusts the pixel value of each pixel in the second equalization area to the corresponding second pixel value to obtain the second area; the computer device multiplies the pixel values of the corresponding pixels in the pre-processing area and the target area to obtain a third pixel value, and adjusts the pixel value of each pixel in the pre-processing area to the corresponding third pixel value to obtain the third area; then the computer device splices the first area, the second area and the third area to obtain the spliced area.

In this embodiment, the first area, the second area and the third area are spliced to obtain the spliced area. Since the spliced area integrates the road features of multiple areas, the features of the road in the spliced area can be made more obvious.

Specifically, the computer device splices the first area, the second area and the third area to obtain the spliced area, which includes: the computer device obtains the first matrix corresponding to the first area, obtains the second matrix corresponding to the second area, and obtains the third matrix corresponding to the third area, and then the computer device splices the first matrix, the second matrix, and the third matrix to obtain the spliced area. The spliced area can be a three-dimensional area.

Step S14, input the spliced area into the pre-trained road segmentation model to obtain a road segmentation image and the segmentation result in the road segmentation image.

In at least one embodiment of the present application, the road segmentation model includes a feature extraction layer, and the feature extraction layer includes a convolution layer, a pooling layer, and a batch normalization layer, etc. The pre-training of the road segmentation model can refer to the training flow chart shown in Figure 5.

Step S141, obtaining a road segmentation network, a road training image, and the annotation results of the road training image.

In this embodiment, the road segmentation network includes, but is not limited to: SegNet, U-Net, FCN and other networks. There are multiple road training images, each of which contains roads. The annotation results in the multiple road training images include multiple annotation pixels corresponding to the roads in each road training image, the annotation positions of the multiple annotation pixels, the annotation quantity of the multiple annotation pixels, and the annotation area formed by the multiple annotation pixels, etc. Among them, the annotation pixels are the pixels corresponding to the roads in each road training image.

Step S142, input the road training image into the road segmentation network to extract features and obtain a road feature map.

In this embodiment, the computer device uses the feature extraction layer to extract features from the road training image to obtain the road feature map.

Step S143, performing road prediction on each pixel point in the road feature map to obtain the prediction result of the road feature map.

In this embodiment, the prediction result includes a plurality of road pixel points corresponding to the road in the road training image, the road positions of the plurality of road pixel points, the number of pixels of the plurality of road pixel points, and the pixel area formed by the plurality of road pixel points, etc.

In this embodiment, the computer device predicts the category of each pixel in each road feature map, obtains multiple preset initial categories for each pixel in each road feature map and the category probability corresponding to each initial category, wherein the multiple initial categories include roads, the computer device determines the initial category corresponding to the maximum category probability as the target category, and the computer device determines the multiple pixel points corresponding to the target category of the road category as the multiple road pixel points, and then the computer device calculates the number of pixels of the multiple road pixel points, obtains the road position of the road pixel points and determines the pixel area formed by the multiple road pixel points. The multiple initial categories can be set by themselves and this application does not limit this. For example, the initial categories include roads, trees, and vehicles, etc.

Step S144, adjusting the parameters of the road segmentation network according to the prediction results and the annotation results to obtain a trained road segmentation model.

In this embodiment, the computer device adjusts the parameters of the road segmentation network according to the prediction result and the annotation result to obtain the trained road segmentation model, including: the computer device calculates the prediction index of the road segmentation network according to the prediction result and the annotation result, and adjusts the parameters of the road segmentation network based on the prediction index until the prediction index meets the preset conditions, thereby obtaining the trained road segmentation model.

Among them, the prediction index can be the prediction accuracy, and the preset condition can be: the prediction accuracy is greater than or equal to the preset valve value or the prediction accuracy no longer increases. The preset valve value can be set by yourself, and this application does not impose any restrictions on this.

Specifically, if the prediction index is the prediction accuracy, the computer device calculates the prediction index of the road segmentation network according to the prediction result and the annotation result, including: the computer device calculates the training quantity of the plurality of road training images, and calculates the prediction quantity of the prediction results that are the same as the corresponding annotation results, and further, the computer device calculates the ratio of the prediction quantity to the training quantity to obtain the prediction accuracy.

The generation process of the road segmentation image is basically the same as the training process of the road segmentation model, so it will not be repeated in this embodiment. As shown in Figure 6, it is a schematic diagram of the road segmentation image provided by the embodiment of this application. The lighter color part in Figure 6 is the road area in the road segmentation image.

In this embodiment, the prediction accuracy is used to determine whether the road segmentation network converges. When the road segmentation network converges, the prediction accuracy is the highest, and the road segmentation model is obtained. Therefore, the detection accuracy of the road segmentation model can be ensured.

In at least one embodiment of the present application, the process of generating the road segmentation image is substantially the same as the process of training the road segmentation model, and the process of generating the segmentation result is substantially the same as the process of generating the prediction result, so the present application will not elaborate on this. Since the road image also needs to be cropped, after obtaining the road segmentation image, the size of the road segmentation image needs to be re-adjusted so that the size of the road segmentation image is the same as the size of the road image.

It can be seen from the above technical solution that the present application performs image processing on the road image to obtain multiple regions of interest. The image processing includes image cropping, binarization and equalization, road detection and Hough transform. Since image cropping can delete the part of the road image that does not contain the road, it can reduce the interference to the road detection. When the road image is a backlit image, by binarizing and equalizing the backlit image, the brightness of the backlit image can be reduced to make the backlit image clearer. By performing road detection on the backlit image, multiple possible initial edge lines of the road in the backlit image can be detected, and the multiple possible edge lines are screened by Hough transform to obtain the target edge line, thereby improving the accuracy of road detection. In addition, since more road information is integrated into the multiple regions of interest, the roads in the spliced regions can be made clearer by generating a spliced region based on the multiple regions of interest. In addition, the roads in the spliced region are detected by using a pre-trained road segmentation model. Since the road segmentation model has learned the location features of the roads, the location of the roads in the spliced region can be predicted more accurately.

As shown in Figure 7, it is a schematic diagram of the structure of the computer device provided in the embodiment of this application.

In one embodiment of the present application, the computer device 1 includes, but is not limited to, a memory 12, a processor 13, and a computer program stored in the memory 12 and executable on the processor 13, such as a road segmentation program.

Those skilled in the art can understand that the schematic diagram is only an example of the computer device 1, and does not constitute a limitation on the computer device 1. The computer device 1 may include more or fewer components than shown in the diagram, or may combine certain components, or may include different components. For example, the computer device 1 may also include input and output devices, model access devices, buses, etc.

The processor 13 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete component gate circuits or transistor components, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 13 is the computing core and control center of the computer device 1, and uses various interfaces and lines to connect various parts of the entire computer device 1, and obtain the operating system of the computer device 1 and various installed applications, program codes, etc.

The processor 13 obtains the operating system of the computer device 1 and various installed applications. The processor 13 obtains the applications to implement the steps in the above-mentioned road segmentation method embodiments, such as the steps shown in Figures 2, 3 and 5.

Exemplarily, the computer program may be divided into one or more modules/units, which are stored in the memory 12 and acquired by the processor 13 to complete the present application. The one or more modules/units may be a series of computer program instruction segments capable of completing a specific function, which are used to describe the acquisition process of the computer program in the computer device 1.

The memory 12 can be used to store the computer program and/or module. The processor 13 realizes various functions of the computer device 1 by running or obtaining the computer program and/or module stored in the memory 12 and calling the data stored in the memory 12. The memory 12 can mainly include a program storage area and a data storage area, wherein the program storage area can store the operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the computer device. In addition, the memory 12 may include a non-volatile memory, such as a hard disk, a memory (memory), a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a memory card (Flash Card), at least one disk storage device, a memory device, or other non-volatile solid-state storage devices.

The memory 12 may be an external memory and/or an internal memory of the computer device 1. Furthermore, the memory 12 may be a physical memory, such as a memory stick, a TF card (Trans-flash Card), etc.

If the module/unit integrated in the computer device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is obtained by the processor, the steps of the above-mentioned method embodiments can be implemented.

The computer program includes computer program code, which may be in source code form, object code form, retrievable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, flash drive, mobile hard drive, magnetic disk, optical disk, computer storage, read-only memory (ROM).

In conjunction with Figure 2, the memory 12 in the computer device 1 stores multiple instructions to implement a road segmentation method, and the processor 13 can obtain the multiple instructions to implement: obtaining a road image; performing image processing on the road image to obtain multiple regions of interest; generating a spliced area based on the multiple regions of interest; inputting the spliced area into a pre-trained road segmentation model to obtain a road segmentation image and a segmentation result in the road segmentation image.

Specifically, the specific implementation method of the processor 13 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of Figure 2, which will not be elaborated here.

In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation.

The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple model units. Some or all of the modules may be selected according to actual needs to achieve the purpose of this embodiment.

In addition, each functional module in each embodiment of the present application may be integrated into a processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes within the meaning and scope of the equivalent elements of the claims in the present application. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

In addition, it is obvious that the word "including" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices described in this application can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of this application and are not limiting. Although this application is described in detail with reference to the preferred embodiments, ordinary technicians in this field should understand that the technical solution of this application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of this application.

S11~S14: Steps

Claims (9)

A road segmentation method is applied to a computer device, wherein the road segmentation method comprises: obtaining a road image, wherein the road image comprises a backlit road image; based on the road image, obtaining a plurality of regions of interest, wherein the plurality of regions of interest comprise a pre-processed region, a first equalized region, a second equalized region, and a target region, wherein the first equalized region is obtained by performing a grayscale bar graph equalization process on the pre-processed region, and the second equalized region is obtained by performing a bar graph equalization process on each channel of the pre-processed region, and the target region is generated by: detecting the first equalized region to obtain a target edge line, and The target edge line and the first equalized area are used to generate the target area; the detecting the first equalized area to obtain the target edge line includes: performing edge detection on the first equalized area to obtain multiple initial edge lines, calculating the polar coordinate equation of each pixel point on each initial edge line based on the pixel position in the first equalized area, and selecting the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations; generating a splicing area based on the multiple regions of interest; inputting the splicing area into a pre-trained road segmentation model to obtain a road segmentation image and a segmentation result in the road segmentation image. The road segmentation method as described in claim 1, wherein obtaining multiple regions of interest based on the road image includes: pre-processing the road image to obtain the pre-processed region. The road segmentation method as described in claim 1, wherein the edge detection of the first equalized area to obtain multiple initial edge lines includes: filtering the first equalized area to obtain a filter area; calculating the gradient value of each pixel in the filter area; filtering multiple pixels corresponding to the multiple gradient values according to a preset upper limit threshold value and a preset lower limit threshold value to obtain edge pixels; Determining multiple lines formed by multiple adjacent edge pixels as the multiple initial edge lines. The road segmentation method as described in claim 1, wherein the target edge line includes a first target edge line and a second target edge line, and the selecting the target edge line from the multiple initial edge lines based on the multiple polar coordinate equations includes: drawing a curve corresponding to each polar coordinate equation, selecting the polar angle of each initial edge line based on the multiple curves; calculating the edge line slope of the corresponding initial edge line according to the polar angle; selecting the initial edge line corresponding to the maximum edge line slope as the first target edge line, and selecting the initial edge line corresponding to the minimum edge line slope as the second target edge line. The road segmentation method as described in claim 4, wherein generating the target area according to the target edge line and the first equalization area includes: determining the area enclosed by the first target edge line and the second target edge line in the first equalization area as the road area; and binarizing the first equalization area based on the road area to obtain the target area. A road segmentation method as described in claim 1, wherein the generation of a spliced area based on the multiple regions of interest includes: multiplying the pixel values of the first equalized area and the corresponding pixel points in the target area to obtain a first pixel value, and adjusting the pixel value of each pixel point in the first equalized area to the corresponding first pixel value to obtain a first area; multiplying the pixel values of the second equalized area and the corresponding pixel points in the target area to obtain The second pixel value is obtained by multiplying the pixel values of the corresponding pixels in the pre-processed area and the target area to obtain a third pixel value, and the pixel value of each pixel in the pre-processed area is adjusted to the corresponding third pixel value to obtain a third area; the first area, the second area and the third area are spliced to obtain the spliced area. The road segmentation method as described in claim 1, wherein before inputting the spliced area into the pre-trained road segmentation model, the method further includes: obtaining a road segmentation network, a road training image, and an annotation result of the road training image; inputting the road training image into the road segmentation network to extract features to obtain a road feature map; performing road prediction on each pixel in the road feature map to obtain a prediction result of the road feature map; adjusting the parameters of the road segmentation network according to the prediction result and the annotation result to obtain a trained road segmentation model. A computer device, wherein the computer device comprises: a memory storing at least one instruction; and a processor executing the at least one instruction to implement the road segmentation method as described in any one of claims 1 to 7. A computer-readable storage medium, wherein at least one instruction is stored in the computer-readable storage medium, and the at least one instruction is executed by a processor in a computer device to implement a road segmentation method as described in any one of claims 1 to 7.