Traffic signal light color recognition method and system

Abstract

The embodiment of the present invention provides a method and system for identifying the color of a traffic light, belonging to the field of image recognition technology. The method comprises: collecting image information of a traffic light scene, and obtaining image information of a traffic light area based on a preset cropping algorithm; performing channel decomposition and recombination on the image information of the traffic light area to obtain recombined image information; performing background difference deletion processing based on the recombined image information to obtain background-deleted image information; and determining the color of the light based on the background-deleted image information. The solution of the present invention can be applied to any image acquisition scenario, and only requires changing the channel combination method, so that a set of traffic light recognition algorithms can be applied to any scenario, thereby improving the universality of the method and facilitating large-scale use.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method for recognizing the color of a traffic light and a system for recognizing the color of a traffic light. Background Art

[0002] Traffic light recognition, directly related to smart transportation, has become a current research hotspot. This has led to a proliferation of methods for identifying the color of traffic lights. While these methods have addressed the issue of traffic light recognition to a certain extent, neither the accuracy nor the universality of these methods have met the requirements for widespread application. For example, as research into intelligent driving becomes increasingly popular, the requirements for traffic light recognition accuracy are also becoming increasingly stringent. For cars to autonomously drive on the road, accurate recognition of traffic lights is essential for subsequent corrective action. Under the fundamental principle of protecting user safety, no recognition errors are permitted. Furthermore, as use becomes increasingly widespread in the future, if each recognition system requires the design of a corresponding recognition algorithm, both development costs and efficiency will be significantly impacted. As scenes become more complex and image pixels and image acquisition equipment become unrestricted, the difficulty of traffic light detection increases dramatically. This difficulty is primarily due to factors such as changing weather conditions and uncontrollable camera-based factors. For example, traffic light recognition based on electronic police (cameras) is a significant challenge in itself. Different types of electronic police in different regions, with varying installation years and equipment aging, further complicate issues such as resolution and color cast. Consequently, traffic light recognition in changing scenes becomes even more challenging. To adapt to traffic light recognition in these changing conditions, a new traffic light recognition method is needed. Summary of the Invention

[0003] The purpose of the embodiments of the present invention is to provide a method and system for identifying the color of a traffic light, so as to at least solve the problem that existing traffic lights cannot adapt to accurate and efficient identification of traffic lights under changing scene conditions.

[0004] In order to achieve the above-mentioned objectives, the first aspect of the present invention provides a method for identifying the color of a traffic light, the method comprising: collecting traffic light scene image information, and obtaining traffic light area image information based on a preset cropping algorithm; performing channel decomposition and recombination on the traffic light area image information to obtain recombined image information; performing background difference deletion processing based on the recombined image information to obtain background-deleted image information; and determining the color of the light based on the background-deleted image information.

[0005] Optionally, obtaining traffic light area image information based on a preset cropping algorithm includes: preprocessing the traffic light scene image information to obtain preprocessed image information; importing the preprocessed image information into a preset traffic light recognition model to output the corresponding traffic light position area; retaining the traffic light area and cropping the remaining positions to obtain traffic light area image information.

[0006] Optionally, the method further includes: if the preset traffic light recognition model fails to identify the location area including the traffic light, cropping the traffic light image information based on a preset image cropping rule; wherein the preset image cropping rule is to retain at least the area within 1 / 3 below the upper border of the traffic light scene image information.

[0007] Optionally, the channel decomposition and recombination of the traffic light area image information includes: channel decomposing the traffic light image information under a preset color space; channel combining the decomposed image information based on a preset color distinction algorithm, so that the obtained combined image information only includes two parts of information: the lighted area and the background area; wherein the channel combination of the decomposed image information is performed based on a preset channel combination rule; the preset color space includes: any one of RGB, ARGYB, CIELAB, CIELCHAB, CIELCHUV, CIELUV, CIEXYZ, CIEXYZ2, CIEXYZ3, CIEXYZ4, HLS, HIS, HSV, I1I2I3, HIS, LMS, YIQ, and YUV.

[0008] Optionally, the method also includes: obtaining preset channel combination rules, including: obtaining historical image information of the current traffic light, and performing channel decomposition on the historical image information; traversing all existing channel combination methods, and outputting all channel combination results; pushing all channel combination results, and recovering the user's annotation results; using the channel combination method corresponding to the user's annotation result as the preset channel combination rule applicable to the current traffic light; wherein the user's annotation result is a channel combination method that is sensitive to red, green and yellow respectively; the channel combination method that is sensitive to red, green and yellow is: after completing the channel combination, the channel combination method that shows the corresponding color area based on grayscale distinction in the image information.

[0009] Optionally, the channel combination of the disassembled image information based on the preset color differentiation algorithm makes the obtained combined image information include only two parts of information: the lighted area and the background area, including: performing image channel combination based on the channel combination method sensitive to red, green and yellow, respectively, to obtain corresponding recognition results; for each recognition result, judging whether there are only two color areas with different grayscales: if there are only two color areas with different grayscales, it means that the current recognition result meets the requirements, then the sensitive color of the channel combination method corresponding to the current recognition result is determined to be the color to be recognized, and the image with only two grayscales is regarded as an image including only two parts of information: the lighted area and the background area; if there are not only two color areas with different grayscales, it means that the current recognition result does not meet the requirements, and the recognition result is directly excluded.

[0010] Optionally, the background differential deletion processing is performed based on the recombined image information, including: in multiple combined image information including only the lighted area and the background area, setting the lighted area as area 1 and the background area as area 0; performing differential deletion of the lighted area and the background area to obtain the image information after the background is deleted; wherein, the image information after the background is deleted only includes the lighted area information.

[0011] Optionally, determining the light color based on the image information after the background is deleted includes: after the background deletion is completed, performing color judgment on the remaining area according to the sensitive color of the current processing image channel combination method; wherein, the remaining area is the light area remaining after the background is deleted; the light color of the light area is the same as the sensitive color of the corresponding image channel combination method.

[0012] A second aspect of the present invention provides a traffic light color recognition system, the system comprising: an acquisition unit for acquiring traffic light scene image information and obtaining traffic light area image information based on a preset cropping algorithm; a processing unit for performing channel decomposition and recombination on the traffic light area image information to obtain recombined image information; a background deletion unit for performing background difference deletion processing based on the recombined image information to obtain background-deleted image information; and an identification unit for determining the light color based on the background-deleted image information.

[0013] On the other hand, the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, enables the computer to execute the above-mentioned method for identifying the color of a traffic light.

[0014] Through the above technical solution, the collected image information is subjected to arbitrary channel combination, and a combination method that is sensitive to the red, green and yellow colors of traffic lights is found. Then, based on this method, the complex background is directly eliminated. Different combination methods can be implemented for different application scenarios, so that the solution of the present invention can be applied to any image acquisition scenario. It only needs to change the channel combination method, so that a set of traffic light recognition algorithms can be applied to any scenario, which improves the universality of the method and is convenient for large-scale use.

[0015] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings:

[0017] Figure 1 This is a flowchart of the steps of a method for identifying the color of a traffic light provided by one embodiment of the present invention;

[0018] Figure 2 This is a system structure diagram of a traffic light color recognition system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0020] Traffic light recognition, directly related to smart transportation, has become a current research hotspot. This has led to a proliferation of methods for identifying the color of traffic lights. While these methods have addressed the issue of traffic light recognition to a certain extent, neither the accuracy nor the universality of these methods have met the requirements for widespread application. For example, as research into intelligent driving becomes increasingly popular, the requirements for traffic light recognition accuracy are also becoming increasingly stringent. Accurately identifying traffic lights is essential for cars to automatically drive on the road, and accurate traffic light recognition is essential for subsequent corrective action. Under the fundamental principle of protecting user safety, no recognition errors are permitted. Furthermore, as use becomes increasingly widespread in the future, if each recognition system requires the design of a corresponding recognition algorithm, both development costs and efficiency will be significantly impacted.

[0021] The primary issue contributing to the lack of universal applicability of current traffic light recognition methods is that they are all developed based on fixed scenarios. For example, they restrict the operating parameters of the image acquisition device, resulting in a relatively fixed image standard for processing. Changes to the image acquisition device or the captured image can render the previously defined algorithm inapplicable, potentially leading to significant deviations in the final recognition results and misidentification. Traditional methods require customization of pixel and camera types. They primarily address complex scenarios, primarily manifested in the diversity of traffic lights, complex lighting conditions, field of view uncertainty, and uncertainty in light brightness. Commonly used methods rely on traditional vision-based grayscale processing, thresholding, dilation and erosion, and top-hat operations, sometimes with color clustering and the fuzzy logic discussed in previous patents. Deep learning-based methods for most attribute-based object detection typically utilize methods such as FastCNN, SSD, and YOLO. Overall, single-camera object detection has largely met industry standards.

[0022] However, when scenes become complex and image pixels and image acquisition devices are unrestricted, the difficulty of traffic light detection increases dramatically. This difficulty is primarily due to complex scenarios such as changing weather conditions, as well as factors uncontrollable by the camera. For example, traffic light recognition based on electronic police (cameras) is inherently challenging. Different types of electronic police in different regions, varying installation ages, and varying equipment aging further complicate resolution and color cast issues, further increasing the difficulty of traffic light recognition based on multiple cameras. The present invention addresses the current problem of poor universality of traffic lights by proposing a novel traffic light color recognition method. The proposed method combines captured image information into arbitrary channels to find a combination that is sensitive to the red, green, and yellow colors of the traffic light. Based on this combination, complex backgrounds are then directly removed. Different combinations can be implemented for different application scenarios, making the present method applicable to any image acquisition scenario. Simply changing the channel combination allows for a single traffic light recognition algorithm to be applied to any scenario, improving the method's universality and facilitating widespread use.

[0023] Figure 2 This is a system structure diagram of a traffic light color recognition system provided by an embodiment of the present invention. Figure 2As shown, an embodiment of the present invention provides a traffic light color recognition system, which includes: an acquisition unit, used to acquire traffic light scene image information, and obtain traffic light area image information based on a preset cropping algorithm; a traffic light color recognition system processing unit, used to perform channel decomposition and recombination on the traffic light area image information to obtain recombined image information; a background deletion unit, used to perform background difference deletion processing based on the recombined image information to obtain background deleted image information; and an identification unit, used to determine the light color based on the background deleted image information.

[0024] Figure 1 This is a flow chart of a method for identifying the color of a traffic light provided by an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a method for identifying the color of a traffic light, the method comprising:

[0025] Step S10: collecting traffic light scene image information, and cropping it based on a preset cropping algorithm to obtain traffic light area image information.

[0026] Specifically, regardless of the traffic light recognition method, all methods process traffic light image information to obtain the color information of the traffic light corresponding to the image. The present invention's solution is similar to conventional methods, both performing specific color recognition based on traffic light image information. Therefore, before the method begins, it is necessary to first capture traffic light image information. This image information can be collected by an electronic police or other pre-defined image acquisition module, such as an image sensor in a self-driving car. After image acquisition is completed, the corresponding image information is sent to a processing unit.

[0027] After the processing unit acquires the current traffic light image information, if it directly identifies the traffic light color in the current image, it will inevitably suffer from reduced processing efficiency and accuracy due to excessive background interference within the scene image. Whether it is an electronic police or a commonly used image acquisition device, the range of the scene image it captures is very large. The traffic light may only be located in a small area within the image information. If the original image information is processed directly, other light information or other interfering information with similar colors to the traffic light may be included, and other lights may be mistakenly identified as traffic lights. Even if there is no other interfering background with similar colors, the need to process such a large area of ​​image information will inevitably affect the processing efficiency of the processing unit. Therefore, image cropping is necessary for both accuracy and processing efficiency.

[0028] The only criterion for image cropping is that the remaining image must include the traffic light. However, different image acquisition devices and image acquisition angles can cause the position coordinates of the traffic light in the image to change, so presetting fixed cropping coordinates is definitely not enough. Based on this, differentiated image cropping is required for different scenarios to ensure that the cropped image information always includes the traffic light.

[0029] In one possible implementation, a traffic light recognition model is proposed. The traffic light recognition model can extract and recognize features of traffic lights in scene images, identify the location of traffic lights, and then perform cropping based on the location by expanding a certain distance in all directions. This can retain only the image information of traffic lights to a great extent. When constructing the traffic light recognition model, a large amount of traffic light image information under the current technical background is collected, including traffic light image information of different traffic light types, different weather conditions, and different time points as training samples. Then, the corresponding model training is performed based on the labeled traffic lights, and the model is tested based on other traffic light images to ensure that the model can recognize traffic lights in images of any weather, any time, and any image acquisition device. As long as the traffic lights can be accurately identified, image cropping based on the recognition results will become simple.

[0030] In another possible embodiment, when the background becomes extremely complex, the traffic light recognition model may not be able to locate the specific position of the traffic light, resulting in the inability to output the corresponding recognition result, and the processing unit will not be able to perform image cropping based on the recognition result. In this case, the solution of the present invention provides a remedy, namely, to crop out useless background areas as much as possible. Although the specific coordinates of the traffic light in the image cannot be fixed, the area where it appears is predictable. Whether it is an electronic police or an image probe, when collecting traffic light images at an intersection, the position of the traffic light must be located in the upper part of the image. If the traffic light image is located in the lower part of the image, it will make it impossible to locate other vehicles, which is inconsistent with the actual application scenario. Furthermore, the biggest obstacle to traffic light recognition is the car taillights, because the color of the car taillights is similar to the color of the traffic light. Therefore, as long as this part of the biggest interference can be removed, the impact on the subsequent accurate recognition of traffic lights will not be significant. Based on this, when the traffic light recognition model is unable to perform silent recognition of traffic lights, it outputs a recognition error signal. In response to the recognition error signal, a remedial cropping solution is executed to crop the car image area in the scene image, retaining only the image information of about 1 / 3 of the upper part of the image, so as to minimize the impact of complex background on subsequent recognition.

[0031] Step S20: performing channel decomposition and recombining on the image information of the traffic light area to obtain recombined image information.

[0032] Specifically, after completing the image cropping, the regional image of the traffic light is obtained. As mentioned above, the solution of the present invention is based on different channel combination rules to perform background removal of the light area. Therefore, before performing channel combination, it is first necessary to perform channel decomposition of the corresponding traffic light regional image according to the preset color system. For example, under the RGB system, it is decomposed into three independent channels of R, G, and B. Different color systems have different corresponding channel decomposition methods. The solution of the present invention is also applicable to other color systems, such as ARGYB, CIELAB, CIELCHAB, CIELCHUV, CIELUV, CIEXYZ, CIEXYZ2, CIEXYZ3, CIEXYZ4, HLS, HIS, HSV, I1I2I3, HIS, LMS, YIQ, and YUV. Then, channel combination is performed based on the decomposed channels. Different channel combinations will highlight different color information. For example, a certain channel combination result will only show a yellow result, and other colors are unified into one color. That means the current combination method is sensitive to yellow. The solution of the present invention is to find the channel combination relationship corresponding to the red, yellow and green colors of traffic lights in the corresponding scenes. Subsequently, image processing can be directly performed based on the channel combination relationship. The directly combined image can highlight the corresponding area, and the color of the area can be directly determined without being affected by other interfering colors.

[0033] Therefore, further, the solution of the present invention also includes finding the channel combination relationship applicable to the current scene. Preferably, the channel combination is disassembled, all existing channel combination methods are traversed, all channel combination results are output, all channel combination results are pushed, and the user's annotation results are recovered. The pushed combination results include a variety of color highlighting relationships. The user selects the correct combination relationship for highlighting the color according to the specific color display of the traffic light, and uses the selected correct combination relationship as the annotation result. After the user's annotation structure is recovered, the channel combination method corresponding to the user's annotation result is used as the channel combination preset rule applicable to the current traffic light. Among them, the user's annotation results are respectively for the red, green and yellow sensitive channel combination methods; the red, green and yellow sensitive channel combination methods are: after the channel combination is completed, only the channel combination method of the corresponding color area is displayed in the image information.

[0034] Specifically, based on the RGB color system distance, the color distinction algorithm trans-from-RGB is used to convert RGB to another color space to distinguish different colors. Based on the input image red channel, input image green channel, and input image blue channel, the corresponding image channels 1, 2, and 3 are output. For example, a channel combination relationship is:

[0035] Img = a*R+b*G+c*B

[0036] Among them, a, b, and c are combination values, and their values ​​range from -1 to 1. Through this combination relationship, the image information is channel-combined. The specific conversion relationship is:

[0037]

[0038] Multiplying this matrix with the three channels preserves the desired color while removing the complex background. This complex background can then be removed using the difference method. Different color systems correspond to different a, b, and c values, but the basic range is limited to -20 to 20. Different values ​​correspond to different background removal capabilities and require adaptive adjustment based on the usage scenario. Other color systems can also use the same channel combination principle, requiring different color differentiation algorithms, which will not be discussed here.

[0039] Step S30: performing background difference deletion processing based on the recombined image information to obtain background deleted image information.

[0040] Specifically, the image channels are combined based on the red, green, and yellow sensitive channel combinations to obtain three recognition results; for each result, it is determined whether there are only two grayscale areas; if there are only two grayscale areas, it means that the current recognition result meets the requirements, then the sensitive color of the channel combination corresponding to the current recognition result is determined to be the color to be recognized, and the image with only two pigments is treated as an image that only includes information of the light area and the background area; if there are not only two grayscale areas, it means that the current recognition result does not meet the requirements, and the recognition result is directly excluded. Continuing with the above RGB channel combination relationship, through image difference, we can obtain the area image we want to retain, and the corresponding difference relationship is:

[0041] ImgeR = (img2 – img3)*3

[0042] ImgeY = (img2 – img3)*3

[0043] ImgeG = (img3 – img2)*3

[0044] Among them, ImgeR, ImgeY, and ImgeG are the processed red, yellow, and green, respectively, which are the colors of the area retained by the difference.

[0045] Step S40: determining the light color based on the image information after background removal.

[0046] Specifically, after the background is deleted, the color of the remaining area is determined according to the sensitive color of the current image channel combination; wherein the remaining area is the lighted area remaining after the background is deleted; the lighted color domain of the lighted area is the same as the sensitive color of the corresponding image channel combination.

[0047] In an embodiment of the present invention, the scheme of the present invention presets different channel combination relationships for different application scenarios, so that the light color of the preset traffic light can be distinguished in the corresponding application scenario based on the preset channel combination relationship, and all other interfering backgrounds are deleted, so that the retained area must be the light area, and the corresponding distinguished color must be the light color of the current traffic light. Under the premise of ensuring the recognition accuracy of the traffic light, the universality of the traffic light recognition method is greatly improved, and it is not affected by changes in image acquisition equipment, time and weather.

[0048] An embodiment of the present invention further provides a computer-readable storage medium, on which instructions are stored, which, when executed on a computer, enable the computer to execute the above-mentioned xx.

[0049] Those skilled in the art will appreciate that all or part of the steps in the methods described in the aforementioned embodiments can be performed by instructing the relevant hardware through a program. The program, stored in a storage medium, includes instructions for causing a microcontroller, chip, or processor to execute all or part of the steps in the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0050] The above describes in detail the optional embodiments of the present invention in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept of the embodiments of the present invention, a variety of simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the scope of protection of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner unless there is any contradiction. In order to avoid unnecessary repetition, the embodiments of the present invention will no longer describe the various possible combinations separately.

[0051] In addition, the various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed in the embodiments of the present invention.

Claims

1. A method for identifying the color of a traffic light, characterized in that: The method comprises: Collect traffic light scene image information and obtain traffic light area image information based on a preset cropping algorithm; performing channel decomposition and recombining on the image information of the traffic light area to obtain recombined image information; Performing background difference deletion processing based on the recombined image information to obtain background deleted image information; Determining the light color based on the image information after the background is deleted; The channel decomposition and recombining of the traffic light area image information includes: Decomposing the traffic light image information into channels in a preset color space; Combining the channels of the disassembled image information based on a preset color distinction algorithm so that the obtained combined image information only includes information of the lighted area and the background area; wherein the channel combination of the disassembled image information is performed based on a preset channel combination rule; Get preset channel combination rules, including: Obtaining historical image information of the current traffic light and performing channel decomposition on the historical image information; Traverse all existing channel combinations and output all channel combination results; Push all channel combination results and retrieve the user's annotation results; The channel combination method corresponding to the user's annotation result is used as the preset channel combination rule applicable to the current traffic light; wherein, The user labeling result is a channel combination method that is sensitive to red, green and yellow respectively; The channel combination method that is sensitive to red, green and yellow is a channel combination method in which corresponding color areas are displayed based on grayscale distinction in image information after the channel combination is completed.

2. The method according to claim 1, characterized in that The obtaining of traffic light area image information based on a preset cropping algorithm includes: Preprocessing the traffic light scene image information to obtain preprocessed image information; Importing the preprocessed image information into a preset traffic light recognition model and outputting the corresponding traffic light location area; The traffic light area is retained, and the remaining positions are cropped to obtain image information of the traffic light area.

3. The method according to claim 2, characterized in that The method further comprises: If the preset traffic light recognition model fails to recognize the location area including the traffic light, the traffic light image information is cropped based on the preset image cropping rules; wherein, The preset image cropping rule is to retain at least the area within 1 / 3 below the upper border of the traffic light scene image information.

4. The method according to claim 1, wherein The preset color space includes: Any of RGB, ARGYB, CIELAB, CIELCHAB, CIELCHUV, CIELUV, CIEXYZ, CIEXYZ2, CIEXYZ3, CIEXYZ4, HLS, HIS, HSV, I1I2I3, HIS, LMS, YIQ, and YUV.

5. The method according to claim 1, characterized in that The channel combination of the image information after disassembly based on the preset color differentiation algorithm is performed so that the obtained combined image information only includes two parts of information: the light area and the background area, including: Image channels are combined based on the channel combination methods that are sensitive to red, green, and yellow to obtain corresponding recognition results; For each recognition result, determine whether there are only two color areas with different grayscales: If there are only two color areas with different grayscales, it means that the current recognition result meets the requirements. Then the sensitive color of the channel combination corresponding to the current recognition result is determined to be the color to be recognized. The image with only two grayscales is regarded as an image that only includes information of the light area and the background area. If there are more than two color areas with different grayscales, it means that the current recognition result does not meet the requirements and the recognition result is directly excluded.

6. The method according to claim 5, characterized in that The background difference removal process based on the reassembled image information includes: In the image information obtained by combining multiple parts including only information of the lighted area and the background area, the lighted area is set as area 1, and the background area is set as area 0; Perform differential deletion of the lighted area and the background area to obtain image information after background deletion; wherein the image information after background deletion only includes information of the lighted area.

7. The method according to claim 1, characterized in that The determining of the light color based on the image information after the background is deleted includes: After the background is removed, the color of the remaining area is determined based on the sensitive color of the current image channel combination; wherein the remaining area is the light area remaining after the background is removed; The lighting color of the lighting area is the same as the sensitive color of the corresponding image channel combination.

8. A traffic light color recognition system, characterized in that: The system comprises: An acquisition unit, configured to acquire image information of a traffic light scene and obtain image information of a traffic light area based on a preset cropping algorithm; a processing unit, configured to perform channel decomposition and recombining on the image information of the traffic light area to obtain recombined image information; A background deletion unit, configured to perform background difference deletion processing based on the recombined image information to obtain background deleted image information; a recognition unit, configured to determine a light color based on the image information after background removal; The channel decomposition and recombining of the traffic light area image information includes: Decomposing the traffic light image information into channels in a preset color space; Combining the channels of the disassembled image information based on a preset color distinction algorithm so that the obtained combined image information only includes information of the lighted area and the background area; wherein the channel combination of the disassembled image information is performed based on a preset channel combination rule; Get preset channel combination rules, including: Obtaining historical image information of the current traffic light and performing channel decomposition on the historical image information; Traverse all existing channel combinations and output all channel combination results; Push all channel combination results and retrieve the user's annotation results; The channel combination method corresponding to the user's annotation result is used as the preset channel combination rule applicable to the current traffic light; wherein, The user labeling result is a channel combination method that is sensitive to red, green and yellow respectively; The channel combination method that is sensitive to red, green and yellow is a channel combination method in which corresponding color areas are displayed based on grayscale distinction in image information after the channel combination is completed.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the method for identifying the color of a traffic light according to any one of claims 1 to 7.

Images