Image screening method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to an image screening method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a scene image shot by a vehicle camera in a vehicle driving process; removing abnormal values and noise in the scene image to obtain a cleaned scene image; based on the trained feature extraction network, identifying the cleaned scene image to obtain an appearing content feature type and a corresponding appearing probability; based on content feature types appearing in the scene images and corresponding appearing probabilities, the suitability between each cleaned scene image and a subsequent processing process is scored; and screening out the cleaned scene image of which the suitability score is greater than a preset threshold value from the cleaned scene images. By adopting the method, the scene image meeting the preset condition can be screened out, a basis is provided for development and model optimization of the automatic driving system, and the automatic driving system can make an accurate decision.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image screening method, apparatus, computer equipment, computer-readable storage medium and computer program product. Background Art

[0002] With the rapid development of autonomous driving technology, the demand for autonomous driving in commercial vehicles is becoming stronger and stronger. The application of artificial intelligence driving makes the safety and reliability of autonomous driving systems largely dependent on the quality of collected data. High-quality data can provide a good foundation for the development and model optimization of autonomous driving systems, ensuring that autonomous driving systems can make accurate decisions in complex environments.

[0003] At present, there is no unified method for image screening of autonomous driving data. The results of image screening are not accurate and lack consistency and comparability, which affects the research and development efficiency of autonomous driving technology and poses a threat to the safety and reliability of autonomous driving systems. Summary of the invention

[0004] Based on this, it is necessary to provide an image screening method, device, computer equipment, computer-readable storage medium and computer program product that can accurately screen target images in response to the above technical problems.

[0005] In a first aspect, the present application provides an image screening method, comprising:

[0006] Acquire scene images captured by a vehicle camera during vehicle driving;

[0007] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0008] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0009] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0010] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0011] In one embodiment, removing outliers and noise in the scene image to obtain a cleaned scene image includes:

[0012] Based on the Laplace operator, the clarity of each scene image is calculated;

[0013] Based on the clarity of each scene image, calculating an average clarity of the plurality of scene images, and calculating a clarity threshold according to the average clarity;

[0014] The scene image whose clarity is greater than the clarity threshold is used as the cleaned scene image.

[0015] In one embodiment, the calculation of the definition of each scene image based on the Laplacian operator includes:

[0016] Calculating the convolution of each pixel block in each scene image based on the Laplacian operator and the pixel value of each pixel block in the scene image; the size of the pixel block corresponds to the dimension of the Laplacian operator;

[0017] The definition of each scene image is calculated according to the convolution of each pixel block in each scene image.

[0018] In one embodiment, calculating the definition of each scene image according to the convolution of each pixel block in each scene image includes:

[0019] The absolute value of the convolution of each pixel block in each scene image is calculated, and the absolute value of the convolution of each pixel block in each scene image is added to obtain the definition of each scene image.

[0020] In one embodiment, the step of identifying the cleaned scene image based on the trained feature extraction network to obtain the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image includes:

[0021] Selecting a preset number of scene images from the cleaned scene images for annotation to obtain annotated scene images;

[0022] The feature extraction network is trained based on the annotated scene image until the loss function of the feature extraction network no longer decreases, thereby obtaining a trained feature extraction network; the loss function is obtained by adding the loss functions respectively calculated by the detection heads corresponding to the road type, traffic density, weather conditions and lighting conditions.

[0023] In one embodiment, scoring the compatibility between each cleaned scene image and a subsequent processing process based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image includes:

[0024] According to the preset score corresponding to each content feature type, the product between the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of occurrence is calculated, and the calculated products are added to obtain the adaptability score of each cleaned scene image; wherein the preset score corresponding to the content feature type is positively correlated with the adaptability between the content feature type and the subsequent processing process.

[0025] In a second aspect, the present application also provides an image screening device, comprising:

[0026] An acquisition module is used to acquire scene images captured by a vehicle camera during vehicle driving;

[0027] A cleaning module, used to remove outliers and noise in the scene image to obtain a cleaned scene image;

[0028] a recognition module, configured to recognize the cleaned scene image based on the trained feature extraction network, and obtain the content feature type and the corresponding probability of occurrence in the scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0029] A scoring module, used for scoring the compatibility between each cleaned scene image and a subsequent processing process based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image;

[0030] The screening module is used to screen out the cleaned scene images whose adaptability scores are greater than a preset threshold from the cleaned scene images.

[0031] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0032] Acquire scene images captured by a vehicle camera during vehicle driving;

[0033] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0034] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0035] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0036] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0037] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0038] Acquire scene images captured by a vehicle camera during vehicle driving;

[0039] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0040] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0041] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0042] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0043] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0044] Acquire scene images captured by a vehicle camera during vehicle driving;

[0045] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0046] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0047] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0048] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0049] The above-mentioned image screening method, device, computer equipment, computer-readable storage medium and computer program product first obtain a scene image taken by a vehicle camera during vehicle driving; remove outliers and noise in the scene image to obtain a cleaned scene image; based on the trained feature extraction network, identify the cleaned scene image to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather conditions or lighting conditions; based on the content feature type and the corresponding probability of occurrence in the scene image, score the adaptability between each cleaned scene image and the subsequent processing process; from the cleaned scene image, screen out the cleaned scene image with a adaptability score greater than a preset threshold. In this way, by inputting the cleaned scene image into the trained feature extraction network, the corresponding content feature type and probability are obtained, and the score is scored, so that the scene image that meets the target image is screened out according to the score, providing a good foundation for the development and model optimization of the autonomous driving system, ensuring that the autonomous driving system can make accurate decisions in a complex environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A diagram showing an application environment of an image screening method in one embodiment;

[0052] Figure 2 A schematic diagram of a process of an image screening method in one embodiment;

[0053] Figure 3 is a structural block diagram of a detection head in a feature recognition network in one embodiment;

[0054] Figure 4 is a structural block diagram of an image screening device in one embodiment;

[0055] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] The image screening method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the scene image data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0058] In an exemplary embodiment, Figure 2 As shown, an image screening method is provided, which is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, including the following steps 202 to 210. Among them:

[0059] Step 202, obtaining a scene image captured by a vehicle camera during vehicle driving.

[0060] In some embodiments, the terminal obtains scene images captured by the vehicle during driving through the vehicle camera.

[0061] Step 204: remove outliers and noise in the scene image to obtain a cleaned scene image.

[0062] In some embodiments, the terminal may filter out scene images that meet the requirements as cleaned scene images by combining the Laplacian operator and the pixels in the scene image.

[0063] Step 206 , based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image.

[0064] Among them, the content feature type includes at least one of road type, traffic density, weather conditions or lighting conditions, and can also be other content features customized by the user, which is not limited in the embodiments of the present application.

[0065] Specifically, the terminal inputs the cleaned scene image into the trained feature extraction network, and the trained feature extraction network outputs the content feature types and corresponding occurrence probabilities in the cleaned scene image through different detection heads.

[0066] Among them, the detection head includes a target detection head, a road type detection head, a weather detection head and a light detection head, which are used to detect the content feature type in the cleaned scene image.

[0067] Step 208 : Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene images, score the compatibility between each cleaned scene image and the subsequent processing process.

[0068] In some embodiments, the terminal scores each cleaned scene image according to the content feature types and corresponding occurrence probabilities in the cleaned scene image. The score is used to measure the compatibility between the cleaned scene image and subsequent processing.

[0069] Step 210 , screening out cleaned scene images having a suitability score greater than a preset threshold from the cleaned scene images.

[0070] In some embodiments, after obtaining the scores of all cleaned scene images, all cleaned scene images with scores greater than a preset threshold are screened out as target images that meet the requirements that the user wants to obtain, wherein the preset threshold is the average of the scores of all cleaned scene images.

[0071] In the above-mentioned image screening method, a scene image captured by a vehicle camera during vehicle driving is obtained; outliers and noise in the scene image are removed to obtain a cleaned scene image; based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather conditions or lighting conditions; based on the content feature type and the corresponding probability of occurrence in the scene image, the adaptability between each cleaned scene image and the subsequent processing process is scored; from the cleaned scene image, the cleaned scene image with a adaptability score greater than a preset threshold is screened out. In this way, by inputting the cleaned scene image into the trained feature extraction network, the corresponding content feature type and probability are obtained, and the score is scored, so that the scene image that meets the target image is screened out according to the score, providing a good foundation for the development and model optimization of the autonomous driving system, and ensuring that the autonomous driving system can make accurate decisions in a complex environment.

[0072] In an exemplary embodiment, outliers and noise in a scene image are removed to obtain a cleaned scene image, including: calculating the clarity of each scene image based on a Laplace operator; calculating the average clarity of multiple scene images based on the clarity of each scene image, and calculating a clarity threshold based on the average clarity; and using a scene image with a clarity greater than the clarity threshold as the cleaned scene image.

[0073] In actual implementation, the terminal calculates the clarity of each scene image based on the Laplace operator. , and then calculate the average clarity of multiple scene images according to the clarity of each scene image. The specific calculation formula is shown in formula (1).

[0074]

[0075] in, is the average resolution, and N is the number of all scene images.

[0076] Furthermore, according to the average clarity of multiple scenes, a clarity threshold is calculated. The specific calculation formula is shown in formula (2).

[0077]

[0078] Where μ is a user-defined coefficient.

[0079] Furthermore, the scene image whose clarity is greater than the clarity threshold is used as the cleaned scene image.

[0080] In the above embodiment, the scene image is processed based on the Laplace operator to make the edges and details of the image clearer, and at the same time, the scene images that do not meet the clarity requirements are screened and removed, so that the final scene image is more consistent with the target image.

[0081] In some exemplary embodiments, the clarity of each scene image is calculated based on the Laplacian operator, including: calculating the convolution of each pixel block in each scene image based on the Laplacian operator and the pixel value of each pixel block in the scene image; the size of the pixel block corresponds to the dimension of the Laplacian operator; and calculating the clarity of each scene image based on the convolution of each pixel block in each scene image.

[0082] In some exemplary embodiments, the clarity of each scene image is calculated based on the convolution of each pixel block in each scene image, including: calculating the absolute value of the convolution of each pixel block in each scene image, and adding the absolute value of the convolution of each pixel block in each scene image to obtain the clarity of each scene image.

[0083] In actual implementation, the convolution of each pixel block in each scene image is calculated based on the Laplacian operator and the pixel value of each pixel block in the scene image. The specific calculation formula is shown in (3), where the size of the pixel block corresponds to the dimension of the Laplacian operator. For example, if the Laplacian operator is a 3*3 matrix, then the pixel block is also a 3*3 pixel point set.

[0084]

[0085] in, is a pixel block, and L is the Laplace operator.

[0086] Furthermore, the absolute value of the convolution of each pixel block in each scene image is added to obtain the clarity of each scene image. The specific calculation formula is shown in formula (4).

[0087]

[0088] In the above embodiment, the absolute value of the convolution of each pixel block in each scene image is added, so that the final result is more comprehensive.

[0089] In the above embodiment, the clarity of the entire scene image is obtained by comprehensively considering multiple pixel values ​​in each scene image, so that the final result is more accurate.

[0090] In some exemplary embodiments, before the cleaned scene image is recognized based on the trained feature extraction network and the content feature types and corresponding probabilities appearing in the cleaned scene image are obtained, the method includes: selecting a preset number of scene images from the cleaned scene images for annotation to obtain annotated scene images; training the feature extraction network based on the annotated scene images until the loss function of the feature extraction network no longer decreases to obtain a trained feature extraction network; the loss function is obtained by adding the loss functions calculated by the detection heads corresponding to the road type, traffic density, weather conditions and lighting conditions.

[0091] In actual implementation, the terminal selects a preset number of scene images from the cleaned scene images for annotation to obtain annotated scene images, inputs the annotated scene images into the feature extraction network for training, and calculates the loss function of the feature extraction network. The specific calculation formula is shown in formula (5).

[0092]

[0093] in, is the loss function of the target detection head corresponding to the traffic density, is the loss function of the road type detection head corresponding to the road type, is the loss function of the weather detection head corresponding to the weather condition, is the loss function of the illumination detection head corresponding to the illumination condition, and its specific distribution is as follows Figure 3 shown.

[0094] Furthermore, the specific calculation formula of the loss function of the target detection head is shown in (6), and the calculation formula of the loss function of the road type detection head, the loss function of the weather detection head and the loss function of the illumination detection head are shown in (7).

[0095]

[0096]

[0097] Where N is the preset number, is the true value of the i-th sample, is the predicted value of the i-th sample, M is the number of categories in the content feature type, is an indicator variable indicating whether the i-th sample belongs to the j-th category. To predict the probability that the i-th sample belongs to the j-th category.

[0098] The feature extraction network is continuously trained, and after the loss function is no longer significantly reduced, the trained loss feature extraction network is obtained.

[0099] In the above embodiment, the feature extraction network is continuously trained through scene images to obtain a trained feature extraction network, so that the final output result is more accurate.

[0100] In some exemplary embodiments, based on the content feature types appearing in the cleaned scene image and the corresponding probability of occurrence, the adaptability of each cleaned scene image to the subsequent processing process is scored, including: according to the preset score corresponding to each content feature type, the product between the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of occurrence is calculated, and the calculated products are added to obtain the adaptability score of each cleaned scene image; wherein the preset score corresponding to the content feature type is positively correlated with the adaptability between the content feature type and the subsequent processing process.

[0101] In some embodiments, as shown in Table 1, a corresponding preset score is set for each content feature type.

[0102]

[0103] Table 1 Specific classification of content feature types and corresponding preset scores

[0104] Calculate the product of the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of occurrence, add the calculated products, and obtain the adaptability score of each cleaned scene image. The specific formula is shown in (8).

[0105]

[0106] Among them, Base represents the preset score corresponding to the content feature type in the scene image, and w is the probability of the content feature type in the scene image, among which the probabilities of high, medium, and low traffic density are all set to 1.

[0107] In the above embodiment, by quantifying the content features in the cleaned scene image into a suitability score, the final result is easier to process.

[0108] To explain the image screening method in the present application in detail, an embodiment is used for illustration below. For example, the present application is described with an image screening method in a specific vehicle autonomous driving scenario.

[0109] First, the terminal uses the vehicle camera to obtain the scene images captured by the vehicle during driving.

[0110] The terminal selects a preset number of scene images from the cleaned scene images for annotation, obtains the annotated scene images, inputs the annotated scene images into the feature extraction network for training, and calculates the loss function of the feature extraction network. The specific calculation formula is shown in formula (5). Furthermore, the specific calculation formula of the loss function of the target detection head is shown in (6), and the calculation formulas of the loss function of the road type detection head, the loss function of the weather detection head, and the loss function of the illumination detection head are shown in (7).

[0111] The feature extraction network is continuously trained, and after the loss function is no longer significantly reduced, the trained loss feature extraction network is obtained.

[0112] According to the Laplacian operator and the pixel value of each pixel block in the scene image, the convolution of each pixel block in each scene image is calculated. The specific calculation formula is shown in (3), where the size of the pixel block corresponds to the dimension of the Laplacian operator. For example, if the Laplacian operator is a 3*3 matrix, then the pixel block is also a 3*3 pixel point set. The Laplacian operator can be specifically .

[0113] Furthermore, the absolute value of the convolution of each pixel block in each scene image is added to obtain the clarity of each scene image. The specific calculation formula is shown in formula (4). Then, based on the clarity of each scene image, the average clarity of multiple scene images is calculated. The specific calculation formula is shown in formula (1). Furthermore, based on the average clarity of multiple scenes, the clarity threshold is calculated. The specific calculation formula is shown in formula (2). Where μ is 0.08. The scene image whose clarity is greater than the clarity threshold is regarded as the cleaned scene image.

[0114] The cleaned scene image is input into the trained feature extraction network, and the trained feature extraction network outputs the content feature types and corresponding occurrence probabilities in the cleaned scene image through different detection heads.

[0115] According to the preset score corresponding to each content feature type, the product between the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of appearance is calculated, and the calculated products are added to obtain the suitability score of each cleaned scene image. All cleaned scene images with scores greater than the average of all cleaned scene image scores are regarded as the target images that meet the requirements that the user wants to obtain.

[0116] This application proposes an image screening method that can effectively screen scene image data that meets the requirements and provide a highly reliable data basis for the development of autonomous driving systems. This not only helps to improve the safety and reliability of autonomous driving technology, but also provides important support for the application of autonomous driving technology in the commercial vehicle industry.

[0117] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiment of the present application also provides an image screening device for implementing the above-mentioned image screening method. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above-mentioned method, so the specific limitations in one or more image screening device embodiments provided below can refer to the limitations on the image screening method above, and will not be repeated here.

[0119] In an exemplary embodiment, Figure 4 As shown, an image screening device is provided, including: an acquisition module 401, a cleaning module 402, a recognition module 403, a scoring module 404 and a screening module 405, wherein:

[0120] An acquisition module is used to acquire scene images captured by a vehicle camera during vehicle driving;

[0121] A cleaning module, used to remove outliers and noise in the scene image to obtain a cleaned scene image;

[0122] a recognition module, configured to recognize the cleaned scene image based on the trained feature extraction network, and obtain the content feature type and the corresponding probability of occurrence in the scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0123] A scoring module, used for scoring the compatibility between each cleaned scene image and a subsequent processing process based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image;

[0124] The screening module is used to screen out the cleaned scene images whose adaptability scores are greater than a preset threshold from the cleaned scene images.

[0125] In some embodiments, the cleaning module is also used to calculate the clarity of each scene image based on the Laplace operator; based on the clarity of each scene image, calculate the average clarity of multiple scene images, and calculate the clarity threshold based on the average clarity; and use the scene image with a clarity greater than the clarity threshold as the cleaned scene image.

[0126] In some embodiments, the device also includes a computing module for computing the convolution of each pixel block in each scene image based on the Laplacian operator and the pixel value of each pixel block in the scene image; the size of the pixel block corresponds to the dimension of the Laplacian operator; and the clarity of each scene image is calculated based on the convolution of each pixel block in each scene image.

[0127] In some embodiments, the above-mentioned calculation module is also used to calculate the absolute value of the convolution of each pixel block in each scene image, and add the absolute value of the convolution of each pixel block in each scene image to obtain the clarity of each scene image.

[0128] In some embodiments, the device also includes a training module for selecting a preset number of scene images from the cleaned scene images for annotation to obtain annotated scene images; training the feature extraction network based on the annotated scene images until the loss function of the feature extraction network no longer decreases to obtain a trained feature extraction network; the loss function is obtained by adding the loss functions calculated by the detection heads corresponding to the road type, traffic density, weather conditions and lighting conditions.

[0129] In some embodiments, the above-mentioned scoring module is also used to calculate the product of the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of occurrence according to the preset score corresponding to each content feature type, and add the calculated products to obtain the adaptability score of each cleaned scene image; wherein the preset score corresponding to the content feature type is positively correlated with the adaptability between the content feature type and the subsequent processing process.

[0130] Each module in the above-mentioned image screening device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0131] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store scene image data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an image screening method is implemented.

[0132] The display unit of the computer device is used to form a visually visible picture, and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball, or touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0133] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0134] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0135] Acquire scene images captured by a vehicle camera during vehicle driving;

[0136] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0137] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0138] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0139] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0140] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0141] Acquire scene images captured by a vehicle camera during vehicle driving;

[0142] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0143] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0144] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0145] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0146] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0147] Acquire scene images captured by a vehicle camera during vehicle driving;

[0148] Removing outliers and noise from the scene image to obtain a cleaned scene image;

[0149] Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition;

[0150] Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process;

[0151] From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

[0152] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0153] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0154] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0155] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An image screening method, characterized in that: The method comprises: Acquire scene images captured by a vehicle camera during vehicle driving; Removing outliers and noise from the scene image to obtain a cleaned scene image; Based on the trained feature extraction network, the cleaned scene image is identified to obtain the content feature type and the corresponding probability of occurrence in the cleaned scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition; Based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, scoring the compatibility between each cleaned scene image and the subsequent processing process; From the cleaned scene images, cleaned scene images having a fitness score greater than a preset threshold are screened out.

2. The method according to claim 1, characterized in that The removing of outliers and noise in the scene image to obtain a cleaned scene image includes: Based on the Laplace operator, the clarity of each scene image is calculated; Based on the clarity of each scene image, calculating an average clarity of the plurality of scene images, and calculating a clarity threshold according to the average clarity; The scene image whose clarity is greater than the clarity threshold is used as the cleaned scene image.

3. The method according to claim 2, characterized in that The method of calculating the definition of each scene image based on the Laplace operator includes: Calculating the convolution of each pixel block in each scene image based on the Laplacian operator and the pixel value of each pixel block in the scene image; the size of the pixel block corresponds to the dimension of the Laplacian operator; The definition of each scene image is calculated according to the convolution of each pixel block in each scene image.

4. The method according to claim 3, characterized in that The calculating the definition of each scene image according to the convolution of each pixel block in each scene image includes: The absolute value of the convolution of each pixel block in each scene image is calculated, and the absolute value of the convolution of each pixel block in each scene image is added to obtain the definition of each scene image.

5. The method according to claim 1, characterized in that Before the cleaned scene image is identified based on the trained feature extraction network to obtain the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image, the method includes: Selecting a preset number of scene images from the cleaned scene images for annotation to obtain annotated scene images; The feature extraction network is trained based on the annotated scene image until the loss function of the feature extraction network no longer decreases, thereby obtaining a trained feature extraction network; the loss function is obtained by adding the loss functions respectively calculated by the detection heads corresponding to the road type, traffic density, weather conditions and lighting conditions.

6. The method according to claim 1, characterized in that Scoring the compatibility between each cleaned scene image and a subsequent processing process based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image includes: According to the preset score corresponding to each content feature type, the product between the preset score corresponding to the content feature type appearing in the scene image and the corresponding probability of occurrence is calculated, and the calculated products are added to obtain the adaptability score of each cleaned scene image; wherein the preset score corresponding to the content feature type is positively correlated with the adaptability between the content feature type and the subsequent processing process.

7. An image screening device, characterized in that: The device comprises: An acquisition module is used to acquire scene images captured by a vehicle camera during vehicle driving; A cleaning module, used to remove outliers and noise in the scene image to obtain a cleaned scene image; a recognition module, configured to recognize the cleaned scene image based on the trained feature extraction network, and obtain the content feature type and the corresponding probability of occurrence in the scene image; the content feature type includes at least one of road type, traffic density, weather condition or lighting condition; A scoring module, used for scoring the compatibility between each cleaned scene image and a subsequent processing process based on the content feature types and corresponding occurrence probabilities appearing in the cleaned scene image; The screening module is used to screen out the cleaned scene images whose adaptability scores are greater than a preset threshold from the cleaned scene images.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images