Lane line image recognition method, image recognition model training method and device

By training the image recognition model and converting it into an integer model, the compatibility problem between the processor and the image recognition model was solved, improving recognition efficiency and accuracy, and enabling automated management.

CN116385991BActive Publication Date: 2026-11-17CHONGQING CHANGAN AUTOMOBILE CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310290221.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-03-22
Publication Date
2026-11-17
Estimated Expiration
2043-03-22

AI Technical Summary

Technical Problem

Incompatibility issues between the processor and the image recognition model in image recognition devices lead to low recognition efficiency and increased recognition time.

Method used

By training the image recognition model to be compatible with the processor and converting it into an integer model for quantized perception training, compatibility and recognition efficiency can be improved.

Benefits of technology

It improves the compatibility between models and processors in image recognition devices, reduces recognition time, improves recognition efficiency and accuracy, and realizes automated management of image recognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116385991B_ABST
    Figure CN116385991B_ABST
Patent Text Reader

Abstract

This application relates to a method for recognizing lane line images, a method and apparatus for training an image recognition model, and belongs to the field of image recognition technology. The method is applied to a target device equipped with a processor and an image recognition model. The method includes: acquiring a road image to be recognized, which includes lane lines; inputting the road image to be recognized into a trained image recognition model and utilizing the processor's processing resources to generate a target road image, which is a road image to be recognized marked with target identifiers used to indicate lane lines; and displaying the target road image. This improves the image recognition efficiency of the device.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of image recognition technology, specifically to a method for recognizing lane line images, a method for training an image recognition model, and an apparatus. Background Technology

[0002] In recent years, with the development of technology, image recognition technology has gradually entered people's lives and has been widely used in industries, transportation, entertainment and other fields. For example, image recognition devices (such as terminals) identify lane lines in road images to provide spatial location information of roads for autonomous driving or driver assistance functions.

[0003] Currently, image recognition devices are equipped with processors and image recognition models. When these devices identify lane lines in road images, they rely on the processor's processing resources and the image recognition model to process the acquired road image and identify the lane lines. However, in this technical solution, there may be compatibility issues between the processor and the image recognition model, increasing the image recognition time and reducing the device's recognition efficiency. Summary of the Invention

[0004] This application provides a method for recognizing lane line images, a method for training an image recognition model, and an apparatus, to at least solve the technical problem of low image recognition efficiency in related technologies. The technical solution of this application is as follows:

[0005] According to a first aspect of this application, a method for recognizing lane line images is provided, applied to a target device. The target device is equipped with a processor and an image recognition model. The method includes: the target device acquiring a road image to be recognized, the road image to be recognized including lane lines; the target device inputting the road image to be recognized into a trained image recognition model and utilizing the processor's processing resources to generate a target road image, the target road image being a road image to be recognized marked with target identifiers used to indicate lane lines; and the trained image recognition model being an integer model. Finally, the target device displays the target road image.

[0006] Based on the aforementioned technical methods, the target device trains an image recognition model, enabling the trained model to be compatible with the processor and fully utilize the processor's processing resources. This improves the compatibility between the image recognition model and the processor, reduces the time required for image recognition, and increases recognition efficiency.

[0007] In one possible implementation, the target device includes: an acquisition module, a hardware module, an algorithm module, and a service module. The hardware module is equipped with a processor, and the algorithm module is equipped with a trained image recognition model. The method for "the target device acquiring a road image to be recognized" includes: the target device acquiring the road image to be recognized through the acquisition module. The method for "the target device inputting the road image to be recognized into the trained image recognition model and calling the processor's processing resources to generate a target road image" includes: the target device sending a first instruction to the algorithm module through the service module, the first instruction instructing the acquisition of the road image to be recognized from the acquisition module and inputting the road image to be recognized into the trained image recognition model. The target device, in response to the first instruction, acquires the road image to be recognized from the acquisition module through the algorithm module and sends a second instruction to the hardware module, the second instruction instructing the calling of the processor's processing resources. The target device, through the algorithm module, receives the processing resources from the processor, inputs the road image to be recognized into the trained image recognition model, and generates the target road image.

[0008] Based on the aforementioned technical means, the target device can realize the processes of image recognition, detection, and transmission through the interaction between various modules, thereby improving the efficiency of image recognition.

[0009] In one possible implementation, the target device further includes a transmission module and a display module, and the target device stores preset parameter thresholds. After the above-described "target device generates target road image," the lane line image recognition method further includes: the target device sending the target road image to the transmission module via the algorithm module; the target device receiving the target road image from the algorithm module via the transmission module and determining whether the image parameters of the target road image are greater than the preset parameter threshold; if the image parameters of the target road image are greater than the preset parameter threshold, the target device sending the target road image to the display module via the transmission module. The above-described "target device displays target road image" method includes: the target device receiving the target road image from the transmission module via the display module and displaying the target road image.

[0010] Based on the aforementioned technical means, the target device can ensure the accuracy of the recognition results displayed by detecting the recognition results output by the trained image recognition model.

[0011] In one possible implementation, the lane line image recognition method further includes: if the image parameters of the target road image are less than or equal to a preset parameter threshold, the target device sends an alarm message to the service module through the transmission module. The alarm message indicates an image recognition anomaly. In response to the alarm message, the target device sends a first instruction to the algorithm module through the service module.

[0012] Based on the aforementioned technical means, if the transmission module determines that the image recognition model's recognition result is abnormal, the transmission module can instruct the algorithm module to re-recognize the image through the service module. This enables automated management of image recognition and improves its efficiency.

[0013] According to a second aspect of this application, a method for training an image recognition model is provided, applied to a target device. The method includes: the target device acquiring a training set, which includes multiple preset road images and preset weights; the preset road images are marked with lane markings. The target device trains the image recognition model based on the multiple preset road images until the model converges, obtaining a first recognition model. The target device converts the first recognition model into an integer model according to the preset weights, obtaining a second recognition model. The target device performs quantized perceptual training on the second recognition model based on the multiple preset road images until the model converges, obtaining a trained image recognition model.

[0014] Based on the aforementioned technical means, by converting the image recognition model into an integer model, the target device can improve the compatibility between the converted image recognition model and the processor, thereby increasing the recognition speed. Furthermore, after converting the image recognition model into an integer model, the target device can perform quantized perceptual training on the converted model to obtain a trained image recognition model. This trained model can compensate for the reduction in recognition accuracy during the conversion process, ensuring that the trained image recognition model can accurately recognize images and improving the accuracy of image recognition.

[0015] According to a third aspect of this application, a lane line image recognition device is provided, applied to a target device. The target device is equipped with a processor and an image recognition model. The lane line image recognition device includes an acquisition unit, a processing unit, and a display unit. The acquisition unit is used to acquire a road image to be recognized, which includes lane lines. The processing unit is used to input the road image to be recognized into a trained image recognition model and call the processor's processing resources to generate a target road image. The target road image is a road image to be recognized marked with target identifiers, which are used to indicate lane lines. The trained image recognition model is an integer model. The display unit is used to display the target road image.

[0016] In one possible implementation, the target device includes: an acquisition module, a hardware module, an algorithm module, and a service module. The hardware module is equipped with a processor, and the algorithm module is equipped with a trained image recognition model. Specifically, the acquisition unit is used to acquire a road image to be recognized through the acquisition module. Specifically, the processing unit is used to send a first instruction to the algorithm module through the service module, instructing the acquisition of the road image to be recognized from the acquisition module and inputting the road image to be recognized into the trained image recognition model. In response to the first instruction, the algorithm module acquires the road image to be recognized from the acquisition module and sends a second instruction to the hardware module, instructing the mobilization of processor processing resources. The algorithm module receives processing resources from the processor, inputs the road image to be recognized into the trained image recognition model, and generates a target road image.

[0017] In one possible implementation, the target device further includes a transmission module and a display module, and the target device stores preset parameter thresholds. The processing unit is further configured to send a target road image to the transmission module via the algorithm module. The processing unit is also configured to receive the target road image from the algorithm module via the transmission module and determine whether the image parameters of the target road image are greater than the preset parameter thresholds. If the image parameters of the target road image are greater than the preset parameter thresholds, the processing unit is further configured to send the target road image to the display module via the transmission module. Specifically, the display unit is configured to receive the target road image from the transmission module via the display module and display the target road image.

[0018] In one possible implementation, the processing unit is further configured to send an alarm message to the service module via the transmission module if the image parameters of the target road image are less than or equal to a preset parameter threshold. The alarm message indicates an image recognition anomaly. The processing unit is also configured to send a first instruction to the algorithm module via the service module in response to the alarm message.

[0019] According to the fourth aspect of this application, a training apparatus for an image recognition model is provided, applied to a target device. The training apparatus for the image recognition model includes an acquisition unit and a processing unit. The acquisition unit is used to acquire a training set, which includes multiple preset road images and preset weights. The preset road images are marked with lane markings. The processing unit is used to train the image recognition model based on the multiple preset road images until the image recognition model converges, obtaining a first recognition model. The processing unit is further used to convert the first recognition model into an integer model according to the preset weights, obtaining a second recognition model. The processing unit is further used to perform quantization-based perceptual training on the second recognition model based on the multiple preset road images until the second recognition model converges, obtaining a trained image recognition model.

[0020] According to a fifth aspect provided in this application, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions. The processor is configured to execute instructions to implement the methods described in the first aspect and any possible embodiments thereof.

[0021] According to a sixth aspect provided in this application, a computer-readable storage medium is provided that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0022] According to the seventh aspect provided in this application, a computer program product is provided, the computer program product including computer instructions that, when executed on an electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0023] Therefore, the above-mentioned technical features of this application have the following beneficial effects:

[0024] (1) By training the image recognition model, the target device can adapt the trained image recognition model to the processor, making full use of the processor's processing resources. In this way, the compatibility between the image recognition model and the processor in the device can be improved, the time for the device to recognize images can be reduced, and the recognition efficiency can be improved.

[0025] (2) The target device can realize the process of image recognition, detection and transmission through the interaction between various modules, thereby improving the efficiency of image recognition.

[0026] (3) By detecting the recognition results output by the trained image recognition model, the target device can ensure the accuracy of the recognition results displayed by the device.

[0027] (4) If the transmission module determines that the recognition result of the image recognition model is abnormal, the transmission module can instruct the algorithm module to re-recognize the image through the service module. In this way, automated management of image recognition can be achieved, and the efficiency of image recognition can be improved.

[0028] (5) By converting the image recognition model into an integer model, the target device can improve the compatibility between the converted image recognition model and the processor, thereby increasing the recognition speed. Furthermore, after the target device converts the image recognition model into an integer model, it can perform quantitative perception training on the converted image recognition model to obtain a trained image recognition model. This trained model can compensate for the reduced recognition accuracy of the model during the conversion process, ensuring that the trained image recognition model can accurately recognize images and improve the accuracy of image recognition.

[0029] It should be noted that the technical effects of any of the implementation methods in aspects three through seven can be found in the technical effects of the corresponding implementation methods in aspects one or two, and will not be repeated here.

[0030] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0031] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0032] Figure 1 This is a schematic diagram of the hardware structure of a device according to an exemplary embodiment;

[0033] Figure 2 This is a schematic diagram of the software structure of a device according to an exemplary embodiment;

[0034] Figure 3 This is a flowchart illustrating a method for recognizing lane line images according to an exemplary embodiment;

[0035] Figure 4 This is a schematic diagram illustrating yet another method for recognizing lane line images according to an exemplary embodiment;

[0036] Figure 5 This is a flowchart illustrating yet another method for recognizing lane line images according to an exemplary embodiment;

[0037] Figure 6 This is a flowchart illustrating yet another method for recognizing lane line images according to an exemplary embodiment;

[0038] Figure 7 This is a flowchart illustrating yet another method for recognizing lane line images according to an exemplary embodiment;

[0039] Figure 8 This is a flowchart illustrating a training method for an image recognition model according to an exemplary embodiment;

[0040] Figure 9 This is a structural block diagram of an image recognition model according to an exemplary embodiment;

[0041] Figure 10 This is a structural block diagram of a downsampling layer according to an exemplary embodiment;

[0042] Figure 11 This is a structural block diagram of a bridging layer according to an exemplary embodiment;

[0043] Figure 12 This is a structural block diagram of an upsampling layer according to an exemplary embodiment;

[0044] Figure 13 This is an example schematic diagram of the LeakyReLU activation function according to an exemplary embodiment;

[0045] Figure 14 This is a flowchart illustrating the target device management application layer according to an exemplary embodiment;

[0046] Figure 15 This is a flowchart illustrating the transmission module sending an image according to an exemplary embodiment;

[0047] Figure 16 This is a flowchart illustrating another method for training an image recognition model according to an exemplary embodiment;

[0048] Figure 17 This is a structural block diagram of a lane line image recognition device according to an exemplary embodiment;

[0049] Figure 18 This is a structural block diagram of a training device for an image recognition model according to an exemplary embodiment;

[0050] Figure 19 This is a structural block diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0051] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0052] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0053] Before providing a detailed description of the lane line image recognition method provided in this application embodiment, the implementation environment and application scenarios of this application embodiment will be introduced first.

[0054] First, the application scenarios of the embodiments of this application will be introduced.

[0055] The lane line image recognition method of this application embodiment is applied to the scenario of recognizing road images. In related technologies, during the process of an image recognition device recognizing lane lines in a road image, the image recognition device needs to process the acquired road image through an image recognition model based on the processor's processing resources to identify the lane lines in the road image.

[0056] In summary, current technical solutions may have compatibility issues between the processor and the image recognition model in image recognition devices, increasing the time required for image recognition and reducing the efficiency of image recognition.

[0057] To address the aforementioned issues, this application provides a method for recognizing lane line images. The target device generates a road image marked with lane lines by inputting a road image to be recognized into a trained image recognition model. The trained image recognition model is an integer-type model, specifically a model that has undergone quantization-based perceptual training after being converted to an integer-type model. In other words, by training the image recognition model, the target device adapts it to the processor, fully utilizing the processor's processing resources. This improves the compatibility between the image recognition model and the processor, reduces the time required for image recognition, and enhances recognition efficiency.

[0058] The implementation environment of the embodiments of this application is described below.

[0059] Figure 1 This is a schematic diagram of the hardware structure of a device provided in an embodiment of this application. Figure 1 As shown, the target device 100 may include: a data acquisition module 101, a hardware module 102, an algorithm module 103, a service module 104, a transmission module 105, and a display module 106.

[0060] The acquisition module 101 (such as a camera) is used to acquire still images or videos. The acquisition module 101 can project the acquired optical image onto a photosensitive element, converting the light signal into an electrical signal. Then, the acquisition module 101 can utilize the processing resources of the hardware module 102 to convert the electrical signal into a digital image signal, and process the digital image signal to obtain images in standard RGB, YUV, or other formats. The photosensitive element can be a charge-coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) phototransistor.

[0061] Hardware module 102 may include one or more processors. For example, hardware module 102 may include a graphics processing unit (GPU), an image signal processor (ISP), and a data processing unit (DPU). These different processors may be independent devices or integrated into one or more hardware modules. Furthermore, hardware module 102 may include a processor with a Field Programmable Gate Array (FPGA) chip. Alternatively, hardware module 102 may include a processor based on an FPGA system.

[0062] Algorithm module 103 may include one or more models (or algorithms). For example, algorithm module 103 may include deep neural network models and image recognition models, etc. Algorithm module 103 can call the processing resources of hardware module 102 to input received information (such as images acquired by acquisition module 101) into the model (or algorithm) and output the calculation results.

[0063] The service module 104 is used to manage the acquisition module 101, hardware module 102, algorithm module 103, transmission module 105 and display module 106, and coordinate the operating status between the modules.

[0064] The transmission module 105 is used to forward the calculation results output by the algorithm module 103 to the display module 106. Furthermore, the transmission module 105 is used to forward still images or videos acquired by the acquisition module 101 to the display module 106. For example, the transmission module 105 compresses an image into JPEG format and sends the compressed image to the display module 106.

[0065] The display module 106 (such as a display screen) may include a display panel. The display module 106 is used to receive information from the transmission module 105 and to present the received information through the display panel.

[0066] It should be noted that, in this embodiment of the application, the acquisition module 101, hardware module 102, algorithm module 103, service module 104, transmission module 105 and display module 106 can communicate via wired / wireless communication.

[0067] In some embodiments, the target device 100 may further include a communication module, and the display module 106 may be deployed on other devices. The transmission module 105 may send the calculation results output by the algorithm module 103 and / or the static images or videos acquired by the acquisition module 101 to the device on which the display module 106 is deployed via the communication module.

[0068] The methods described in the following embodiments can all be implemented in the target device 100 having the above-described hardware structure.

[0069] The software system of the target device 100 can adopt a layered architecture, event-driven architecture, microkernel architecture, microservice architecture, or cloud architecture. This application embodiment uses a layered architecture system as an example to exemplify the software structure of the target device 100.

[0070] Figure 2 This is a schematic diagram of the software structure of a device provided in an embodiment of this application. A layered architecture divides the software into several layers, each with a clear role and function. Layers communicate with each other through software interfaces. In some embodiments, the system is divided into four layers, from top to bottom: the application layer, the data processing layer, the protocol parsing layer, and the transmission control protocol (TCP) / internet protocol (IP) layer.

[0071] The application layer can include a series of threads. For example... Figure 2 As shown, the application layer may include: main thread, system scheduling thread, image preprocessing thread, algorithm scheduling thread, image postprocessing thread, and image output thread.

[0072] The main thread is used to establish the TCP connection between the transmission module and the display module, and to manage the system scheduling thread, image preprocessing thread, algorithm scheduling thread, image post-processing thread, and image output thread. For example, the main thread can create the system scheduling thread, image preprocessing thread, algorithm scheduling thread, image post-processing thread, and image output thread.

[0073] The system scheduling thread is used for image processing, algorithm calls, and communication interface calls.

[0074] The image preprocessing thread is used to receive images acquired by the acquisition module and convert them into the image format required by the image recognition model.

[0075] The algorithm scheduling thread is used to schedule image recognition models and input the images output by the image preprocessing thread into the image recognition models. In this embodiment, the algorithm scheduling thread can schedule multiple image recognition models for computation simultaneously.

[0076] The image post-processing thread is used to label the recognition results in the image. In the embodiments of this application, the image post-processing thread can use different colors and shapes of markers to label different types of recognition results.

[0077] The image output thread is used to send images acquired by the acquisition module and / or labeled images output by the image post-processing thread to the display module.

[0078] The data processing layer may include: an online processing thread, a heartbeat processing thread, a sending processing thread, and a configuration file parsing thread.

[0079] The online processing thread is used to send an online packet to the display module, instructing the display module to enter the running state. Furthermore, the online processing thread is also used to detect the running status of the display module by whether or not an online response packet is received from the display module.

[0080] The heartbeat processing thread is used to send heartbeat packets to the display module at preset times, instructing the display module to send back heartbeat response packets, and maintaining a long-term TCP link between the display module and the transmission module.

[0081] The sending processing thread is used to instruct the transmission module to send the images acquired by the acquisition module and / or the labeled images output by the image post-processing thread to the display module.

[0082] The configuration parsing thread is used to parse the initialization (ini) configuration file, which includes configuration information such as the IP address and communication port of the display module.

[0083] The protocol parsing layer may include: threads for parsing and assembling online packets, heartbeat packets, transmit packets, and command packets. These threads enable cyclic redundancy checks (CRC) on protocol packets (such as online packets, heartbeat packets, transmit packets, or command packets), ensuring the correct acquisition of the protocol packets.

[0084] The TCP / IP layer can include: a TCP packet receiving thread, a TCP packet sending thread, and a packet processing thread. The TCP packet is the processed protocol packet.

[0085] The TCP packet receiving thread is used to instruct the transmission module to receive TCP packets from the display module.

[0086] The TCP packet sending thread is used to instruct the transmission module to send TCP packets to the display module.

[0087] The packet processing thread is used to process multiple TCP packets from the display module.

[0088] The methods described in the following embodiments can all be implemented in the target device 100 having the above-described system architecture.

[0089] It should be noted that in some embodiments, the target device 100 can be a server, or the target device 100 can be a terminal.

[0090] The terminal can be a mobile phone, tablet computer, desktop computer, laptop computer, handheld computer, notebook computer, ultra-mobile personal computer (UMPC), netbook, or other device with shooting and processing functions. This application does not impose any special restrictions on the specific form of the terminal. It can interact with the user through one or more methods such as keyboard, touchpad, touch screen, remote control, voice interaction, or handwriting device.

[0091] A server can be a single physical server, or a server cluster consisting of multiple servers. Alternatively, the server cluster can be a distributed cluster. Alternatively, the server can be a cloud server. This application does not limit the specific implementation of the server.

[0092] For ease of understanding, the following is combined with Figures 3-7 The method for recognizing lane line images provided in this application is described in detail.

[0093] Figure 3 This is a flowchart illustrating a lane line image recognition method according to an exemplary embodiment, such as... Figure 3 As shown, the target device in this application embodiment may include: a data acquisition module, a hardware module, an algorithm module, a service module, a transmission module, and a display module.

[0094] It should be noted that, in the following embodiments, the interaction between various modules in the target device is taken as an example to introduce the embodiments of this application. The lane line image recognition method includes the following steps: S301-S314.

[0095] S301, The acquisition module acquires images of the road to be identified.

[0096] The road image to be identified may include lane lines.

[0097] S302, The acquisition module sends the first message to the service module.

[0098] The first message indicates that an image of the road to be identified has been acquired.

[0099] In some embodiments, the service module may receive a first message from the acquisition module and, in response to the first message, execute S303.

[0100] S303, The service module sends the first instruction to the algorithm module.

[0101] The first instruction is used to instruct the acquisition module to obtain the road image to be identified and input the road image to be identified into the trained image recognition model.

[0102] In this embodiment of the application, the algorithm module is equipped with a trained image recognition model, and the trained image recognition module is an integer model.

[0103] It should be noted that, in the embodiments of this application, the image recognition model can be a U-shaped network (U-Net) model.

[0104] In some embodiments, the algorithm module may receive a first instruction from the service module and, in response to the first instruction, execute S304 and S305.

[0105] S304. The algorithm module acquires the road image to be identified from the acquisition module.

[0106] S305, the algorithm module sends a second instruction to the hardware module.

[0107] The second instruction is used to instruct the processor to access its processing resources.

[0108] In this embodiment of the application, the hardware module is equipped with a processor.

[0109] It should be noted that the order of S304 and S305 is not limited in the embodiments of this application. For example, the algorithm module may execute S304 first and then S305. Alternatively, the algorithm module may execute S305 first and then S304. Or, the algorithm module may execute S304 and S305 simultaneously.

[0110] In some embodiments, the hardware module may receive a second instruction from the algorithm module and, in response to the second instruction, execute S306.

[0111] S306, the hardware module provides processor resources to the algorithm module.

[0112] In some embodiments, the algorithm module may receive processing resources from the processor and, in response to the processor's processing resources, execute S307.

[0113] S307 The algorithm module inputs the road image to be recognized into the trained image recognition model to generate the target road image.

[0114] The target road image is a road image to be identified marked with target identifiers, which are used to indicate lane lines.

[0115] In some embodiments, the algorithm module may execute S308 in response to the generated target road image.

[0116] S308, The algorithm module sends the target road image to the transmission module.

[0117] S309. The transmission module receives the target road image from the algorithm module.

[0118] S310, The transmission module determines the image parameters of the target road image.

[0119] In some embodiments, in order to detect a target road image, after determining the image parameters of the target road image, the transmission module may execute S311.

[0120] S311. The transmission module determines whether the image parameters of the target road image are greater than the preset parameter threshold.

[0121] It should be noted that the image parameters in this application are not limited. For example, the image parameter can be resolution. Another example is saturation. Yet another example is image size (or data length).

[0122] In some embodiments, if the transmission module determines that the image parameters of the target road image are greater than a preset parameter threshold, the transmission module executes S312.

[0123] S312, The transmission module sends the target road image to the display module.

[0124] In some embodiments, the display module may receive a target road image from the transmission module and execute S313 in response to the target road image.

[0125] S313, The display module displays the image of the target road.

[0126] In other embodiments, if the transmission module determines that the image parameters of the target road image are less than or equal to a preset parameter threshold, the transmission module does not execute S312, but executes S314.

[0127] S314. The transmission module sends an alarm message to the service module.

[0128] Alarm messages are used to indicate image recognition anomalies.

[0129] In some embodiments, the service module may receive an alarm message from the transmission module and, in response to the alarm message, resend the first instruction to the algorithm module (i.e., re-execute S303).

[0130] In other words, if the transmission module determines that the image recognition model's recognition result is abnormal, the transmission module can instruct the algorithm module to re-recognize the image through the service module. This enables automated management of image recognition and improves its efficiency.

[0131] The technical solution provided by the above embodiments brings at least the following beneficial effects: The target device can acquire the road image to be identified through the acquisition module, and based on the calling function of the service module, the target device can input the road image to be identified into the trained image recognition model in the algorithm module by calling the processing resources of the hardware module to obtain the target road image. That is, by training the image recognition model, the target device makes the trained image recognition model compatible with the processor, making full use of the processor's processing resources. In this way, the compatibility between the image recognition model in the device and the processor can be improved, the time for the device to recognize images can be reduced, and the recognition efficiency can be improved. Furthermore, the target device can determine whether the image parameters of the target road image are greater than a preset parameter threshold through the transmission module. If the image parameters of the target road image are greater than the preset parameter threshold, the target device displays the target road image through the display module. In this way, by detecting the recognition results output by the trained image recognition model, the target device can ensure the accuracy of the recognition results displayed by the device. Moreover, the target device can realize the processes of image recognition, detection, and transmission through the interaction between various modules, thereby improving the efficiency of image recognition.

[0132] In some embodiments, in order to detect the operating status of the display module, such as Figure 4 As shown, before the transmission module sends the target road image to the display module (i.e., S312), the lane line image recognition method provided in this application embodiment further includes the following steps: S401-S402.

[0133] S401, The transmission module sends a third instruction to the display module.

[0134] The third instruction is used to indicate that the system is in an online state and to send an online response message back to the transmission module.

[0135] S402. The transmission module determines whether it has received an online response message from the display module.

[0136] In some embodiments, if the transmission module determines that it has not received an online response message from the display module, the transmission module does not execute S312.

[0137] In other embodiments, if the transmission module determines that it has received an online response message from the display module, the transmission module executes S312.

[0138] It is understandable that the target device can send a third command to the display module through the transmission module, which can wake up the display module and enable the target device to display the target road image through the display module.

[0139] In some embodiments, the online response message may include the address information and port information of the display module. After the transmission module determines that it has received the online response message from the display module, it can parse the online response message to obtain the address information and port information of the display module. Then, based on the address information and port information of the display module, the transmission module can establish a TCP channel with the display module, and then send the target road image to the display module based on the TCP channel between the transmission module and the display module.

[0140] In some embodiments, in order to detect the connection status between the transmission module and the display module, such as Figure 5 As shown, before the transmission module sends the target road image to the display module (i.e., S312), the lane line image recognition method provided in this application embodiment further includes the following steps: S501-S502.

[0141] S501, The transmission module sends the fourth instruction to the display module.

[0142] The fourth instruction is used to instruct the sending of an online response message at a preset time.

[0143] S502, The transmission module determines whether it receives an online response message from the display module at a preset time.

[0144] In some embodiments, if the transmission module determines that it has received an online response message from the display module at a preset time, the transmission module executes S312.

[0145] In other embodiments, if the transmission module determines that it has not received an online response message from the display module at a preset time, the transmission module disconnects the TCP channel with the display module and resends the third instruction to the display module (i.e., re-executes S401).

[0146] Understandably, the target device sends a fourth instruction to the display module through the transmission module, which can detect the connection status between the transmission module and the display module, ensuring that the transmission module and the display module maintain a normal connection, so that the target device can display the target road image through the display module.

[0147] In some embodiments, to ensure that the display module can receive the target road image, such as Figure 6 As shown, after the transmission module sends the target road image to the display module (i.e., S312), the lane line image recognition method provided in this application embodiment further includes the following steps: S601-S604.

[0148] S601, The display module determines whether it has received the target road image from the transmission module.

[0149] In some embodiments, if the display module determines that it has received a target road image from the transmission module, the display module executes S313.

[0150] In other embodiments, if the display module determines that it has not received a target road image from the transmission module, the display module executes S602.

[0151] S602, The display module sends the fifth instruction to the transmission module.

[0152] The fifth instruction is used to instruct the display module to send target road information.

[0153] In some embodiments, the transmission module may receive a fifth instruction from the display module and, in response to the fifth instruction, execute S603.

[0154] S603. The transmission module determines whether the number of fifth instructions from the display module within a preset time period is greater than a preset number threshold.

[0155] In some embodiments, if the transmission module determines that the number of fifth instructions from the display module within a preset time period is greater than a preset number threshold, the transmission module disconnects the TCP channel with the display module and resends the third instruction to the display module (i.e., re-executes S401).

[0156] In other embodiments, if the transmission module determines that the number of fifth instructions from the display module within a preset time period is less than or equal to a preset number threshold, the transmission module executes S604.

[0157] S604. The transmission module sends the target road image to the display module.

[0158] Understandably, the target device can determine whether it has received the target road image through the display module. If not, the target device sends a fifth instruction to the transmission module through the display module, instructing the transmission module to retransmit the target road image to ensure that the display module can receive it. Furthermore, the target device can use the transmission module to determine whether the number of fifth instructions within a preset time period exceeds a preset threshold, and to check the connection status between the transmission module and the display module, ensuring that they maintain a normal connection so that the target device can display the target road image through the display module.

[0159] It should be noted that in the following embodiments, the target device is used as an example to describe the lane line image recognition method provided in this application. Figure 7 As shown, the method for recognizing the lane line image may include: S701-S703.

[0160] S701, The target device acquires an image of the road to be identified.

[0161] It should be noted that the process of the target device acquiring the road image to be identified can be referred to the description in S301 above, and will not be repeated here.

[0162] S702. The target device inputs the image of the road to be identified into the trained image recognition model and calls the processor's processing resources to generate the target road image.

[0163] It should be noted that the process by which the target device inputs the image of the road to be identified into the trained image recognition model and calls the processor's processing resources to generate the target road image can be referred to the descriptions in S302-S307 above, and will not be repeated here.

[0164] S703, The target device displays an image of the target road.

[0165] It should be noted that the process of displaying the target road image on the target device can be referred to the description in S312-S313 above, and will not be repeated here.

[0166] Figure 8 This is a flowchart illustrating a training method for an image recognition model provided in an embodiment of this application. Figure 8 As shown, the method may include: S801-S804.

[0167] S801, Target device acquires training set.

[0168] The training set includes multiple preset road images and preset weights.

[0169] S802. The target device trains an image recognition model based on multiple preset road images until the image recognition model converges, thus obtaining the first recognition model.

[0170] It should be noted that the process of training an image recognition model based on multiple preset road images for the target device until the image recognition model converges and the first recognition model is obtained can be referred to the introduction of training image recognition models in conventional technologies, and will not be elaborated here.

[0171] S803. The target device converts the first recognition model into an integer model according to the preset weights to obtain the second recognition model.

[0172] It should be noted that the process by which the target device converts the first recognition model into an integer model according to the preset weights to obtain the second recognition model can be referred to in the introduction of converting a floating-point model into an integer model in conventional technology, and will not be elaborated here.

[0173] S804. The target device performs quantitative perception training on the second recognition model based on multiple preset road images until the second recognition model converges, thus obtaining the trained image recognition model.

[0174] It should be noted that the process of the target device performing quantitative perception training on the second recognition model based on multiple preset road images until the second recognition model converges and the trained image recognition model is described in the introduction of quantitative perception training in conventional technology, and will not be elaborated here.

[0175] It should be noted that, as Figure 9 As shown, the structure of the trained image recognition model includes: one input layer, four downsampling layers, one bridge layer, four upsampling layers, two outconv layers, and one output layer.

[0176] The downsampling layer is a feature extraction network: it consists of Conv (convolution), BatchNorm (batch normalization), LeakyReLU (activation function), and pooling, and its structure is as follows. Figure 10 As shown.

[0177] The bridging layer consists of Conv (convolution), BatchNorm (batch normalization), and LeakyReLU (activation function), and its structure is as follows: Figure 11 As shown.

[0178] The upsampling layer (up) is a feature fusion network: it connects the upsampled feature map to the feature map on the left. The upsampling layer allows a low-resolution image containing high-level abstract features to be transformed into a high-resolution image while retaining these high-level abstract features. This high-resolution image is then connected to the high-resolution image containing low-level surface features on the left. It uses ConvTranspose (transposed convolution), Conv (convolution), BatchNorm (batch normalization), and LeakyReLU (activation function), and its structure is as follows: Figure 12 As shown.

[0179] Finally, the feature map is generated by two more outconv convolution operations.

[0180] To better address the vanishing gradient problem in the model, the trained image recognition model uses the LeakyReLU activation function, whose mathematical expression is Equation 1:

[0181] y = max(0, x) + α·min(0, x) Formula 1.

[0182] Here, α is a very small constant that preserves some values ​​of the negative axis, ensuring that the feature information of the negative axis is not completely lost, thus solving the gradient vanishing problem. The LeakyReLU activation function graph is shown below. Figure 13 As shown.

[0183] Understandably, by converting the image recognition model to an integer model, the target device can improve the compatibility between the converted model and the processor, thereby increasing recognition speed. Furthermore, after converting the image recognition model to an integer model, the target device can perform quantized perceptual training on the converted model to obtain a trained image recognition model. This trained model can compensate for the reduction in recognition accuracy during the conversion process, ensuring that the trained model can accurately recognize images and improving overall image recognition accuracy.

[0184] The following describes the process of the target device management application layer in the lane line image recognition method provided in this application, with reference to specific embodiments. Figure 14 As shown, the target device management application layer method includes: S1401-S1412.

[0185] S1401. The target device initializes the main thread context.

[0186] S1402. The target device creates a pipe communication file descriptor.

[0187] S1403, Target device initialization data processing layer.

[0188] S1404, The target device initializes the communication context.

[0189] S1405, The target device initiates a TCP connection.

[0190] S1406. The target device determines whether the TCP connection is successful.

[0191] In some embodiments, if the target device determines that the TCP connection has failed, the target device executes S1407.

[0192] S1407, Target device waits 30 seconds.

[0193] In some embodiments, after the target device executes S1407, the target device may re-execute S1405.

[0194] In other embodiments, if the target device determines that the TCP connection is successful, the target device executes S1408.

[0195] S1408, Target device creates system scheduling thread.

[0196] S1409, The target device creates an image preprocessing thread.

[0197] S1410, Target device creation algorithm scheduling thread.

[0198] S1411, The target device creates an image post-processing thread.

[0199] S1412, The target device creates an image output thread.

[0200] The following describes the process of the transmission module sending images in the lane line image recognition method provided in this application, with reference to specific embodiments. For example... Figure 15 As shown, the method for the transmission module to send images includes: S1501-S1508.

[0201] S1501, The transmission module is blocking the reading of pipeline data.

[0202] S1502, The transmission module determines whether pipeline data has arrived.

[0203] In some embodiments, if the transmission module determines that no pipeline data has arrived, the transmission module re-executes S1501.

[0204] In other embodiments, if the transmission module determines that pipeline data has arrived, the transmission module receives the pipeline data and executes S1503.

[0205] S1503, The transmission module determines whether the length of the received data is 0.

[0206] In some embodiments, if the transmission module determines that the length of the received data is 0, the transmission module executes S1504.

[0207] S1504, The transmission module performs error handling.

[0208] In some embodiments, after the transmission module executes S1504, the transmission module may re-execute S1501.

[0209] In some other embodiments, if the transmission module determines that the length of the received data is not 0, the transmission module executes S1505.

[0210] S1505, The transmission module sends pipeline data.

[0211] S1506, The transmission module determines whether the transmission was successful.

[0212] In some embodiments, if the transmission module determines that the transmission was successful, the transmission module re-executes S1501.

[0213] In some other embodiments, if the transmission module determines that the transmission was unsuccessful, the transmission module executes S1507.

[0214] S1507, The transmission module determines whether the number of transmissions is greater than 2.

[0215] In some embodiments, if the transmission module determines that the number of transmissions is less than or equal to 2, the transmission module re-executes S1505.

[0216] In some other embodiments, if the transmission module determines that the number of transmissions is greater than 2, the transmission module executes S1508.

[0217] S1508, The transmission module resets the TCP connection flag.

[0218] In some embodiments, after the transmission module executes S1508, the transmission module may re-execute S1501.

[0219] The following describes the flowchart of the training method for the image recognition model provided in this application, with reference to specific embodiments. Figure 16 As shown, the training methods for the image recognition model include: S1601-S1605.

[0220] S1601, The target device preprocesses the model (i.e., the target device executes S802).

[0221] S1602. The target device transforms the preprocessed model based on the quantization weights to obtain an integer model (i.e., the target device executes S803).

[0222] S1603. The target device performs quantitative perception training on the converted model based on the calibration dataset (labeled data) to obtain the trained model (i.e., the target device executes S804).

[0223] S1604. The target device converts the trained model into a DPU file.

[0224] S1605. The target device deploys the trained model on the processor based on the converted DPU file.

[0225] The above primarily describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the lane line image recognition device or electronic device includes corresponding hardware structures and / or software modules for performing each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0226] This application embodiment can, based on the above method, exemplarily divide a lane line image recognition device or electronic device into functional modules. For example, the lane line image recognition device or electronic device may include functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.

[0227] Figure 17 This is a structural block diagram illustrating a lane line image recognition device according to an exemplary embodiment. (Refer to...) Figure 17 The lane line image recognition device is applied to a target device, which is equipped with a processor and an image recognition model. The lane line image recognition device includes an acquisition unit 1701, a processing unit 1702, and a display unit 1703.

[0228] The acquisition unit 1701 is used to acquire a road image to be recognized, which includes lane lines. The processing unit 1702 is used to input the road image to be recognized into a trained image recognition model and utilize the processor's processing resources to generate a target road image. The target road image is a road image to be recognized marked with target identifiers, which are used to indicate lane lines. The trained image recognition model is an integer model. The display unit 1703 is used to display the target road image.

[0229] In one possible implementation, the target device includes: an acquisition module, a hardware module, an algorithm module, and a service module. The hardware module is equipped with a processor, and the algorithm module is equipped with a trained image recognition model. The acquisition unit 1701 is specifically used to acquire a road image to be recognized through the acquisition module. The processing unit 1702 is specifically used to send a first instruction to the algorithm module through the service module. The first instruction instructs the acquisition of the road image to be recognized from the acquisition module and inputs the road image to be recognized into the trained image recognition model. In response to the first instruction, the algorithm module acquires the road image to be recognized from the acquisition module and sends a second instruction to the hardware module. The second instruction instructs the mobilization of the processor's processing resources. The algorithm module receives the processing resources from the processor, inputs the road image to be recognized into the trained image recognition model, and generates a target road image.

[0230] In one possible implementation, the target device further includes a transmission module and a display module, and the target device stores a preset parameter threshold. The processing unit 1702 is further configured to send a target road image to the transmission module via the algorithm module. The processing unit 1702 is also configured to receive a target road image from the algorithm module via the transmission module and determine whether the image parameters of the target road image are greater than the preset parameter threshold. The processing unit 1702 is further configured to, if the image parameters of the target road image are greater than the preset parameter threshold, send the target road image to the display module via the transmission module. The display unit 1703 is specifically configured to receive the target road image from the transmission module via the display module and display the target road image.

[0231] In one possible implementation, the processing unit 1702 is further configured to send an alarm message to the service module via the transmission module if the image parameters of the target road image are less than or equal to a preset parameter threshold. The alarm message indicates an image recognition anomaly. The processing unit 1702 is also configured to send a first instruction to the algorithm module via the service module in response to the alarm message.

[0232] Figure 18 This is a structural block diagram illustrating a training apparatus for an image recognition model according to an exemplary embodiment. (Refer to...) Figure 18 The training device for the image recognition model is applied to the target device. The training device for the image recognition model includes an acquisition unit 1801 and a processing unit 1802.

[0233] Acquisition unit 1801 is used to acquire a training set, which includes multiple preset road images and preset weights. The preset road images are marked with lane markings. Processing unit 1802 is used to train an image recognition model based on the multiple preset road images until the image recognition model converges, obtaining a first recognition model. Processing unit 1802 is also used to convert the first recognition model into an integer model according to the preset weights, obtaining a second recognition model. Processing unit 1802 is also used to perform quantized perceptual training on the second recognition model based on the multiple preset road images until the second recognition model converges, obtaining a trained image recognition model.

[0234] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0235] Figure 19 This is a structural block diagram of an electronic device according to an exemplary embodiment. For example... Figure 19 As shown, the electronic device 1900 includes, but is not limited to, a processor 1901 and a memory 1902.

[0236] The aforementioned memory 1902 is used to store the executable instructions of the aforementioned processor 1901. It is understood that the aforementioned processor 1901 is configured to execute instructions to implement the lane line image recognition method in the above embodiment.

[0237] It should be noted that those skilled in the art will understand that Figure 19 The electronic device structure shown does not constitute a limitation on the electronic device; the electronic device may include, but is not limited to, other electronic devices. Figure 19 This may indicate more or fewer components, or combinations of certain components, or different component arrangements.

[0238] Processor 1901 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in memory 1902, and by calling data stored in memory 1902, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Processor 1901 may include one or more processing units. Optionally, processor 1901 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 1901.

[0239] The memory 1902 can be used to store software programs and various data. The memory 1902 may primarily include a program storage area and a data storage area, wherein the program storage area may store the operating system, application programs (such as processing units) required by at least one functional module, etc. Furthermore, the memory 1902 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0240] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 1902 including instructions, which can be executed by a processor 1901 of an electronic device 1900 to implement the lane line image recognition method in the above embodiments.

[0241] In actual implementation, Figure 17 The functions of the acquisition unit 1701, processing unit 1702, and display unit 1703 in the process and Figure 18 The functions of the acquisition unit 1801 and the processing unit 1802 can both be provided by Figure 19 The processor 1901 calls the computer program stored in the memory 1902 to implement the process. The specific execution process can be found in the description of the lane line image recognition method in the previous embodiment, and will not be repeated here.

[0242] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0243] In an exemplary embodiment, this application also provides a vehicle, the vehicle including Figure 19 The processor 1901 shown is used in this vehicle to perform the lane line image recognition method described in the above embodiments.

[0244] In an exemplary embodiment, this application also provides a computer program product including one or more instructions, which can be executed by a processor of an electronic device to complete the lane line image recognition method in the above embodiments.

[0245] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of an electronic device, they implement the various processes of the lane line image recognition method embodiment described above, and can achieve the same technical effect as the lane line image recognition method described above. To avoid repetition, they will not be described again here.

[0246] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0247] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0248] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0249] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0250] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0251] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for recognizing lane line images, characterized in that, The method is applied to a target device, which is equipped with a processor and an image recognition model. The target device includes: an acquisition module, a hardware module, an algorithm module, a service module, a transmission module, and a display module. The hardware module is equipped with the processor, and the algorithm module is equipped with a trained image recognition model. The target device stores preset parameter thresholds. The method includes: The acquisition module acquires an image of the road to be identified, the image of the road to be identified including lane lines; The service module sends a first instruction to the algorithm module, the first instruction being used to instruct the acquisition module to obtain the road image to be identified and input the road image to be identified into the trained image recognition model. The algorithm module responds to the first instruction by acquiring the road image to be identified from the acquisition module and sending a second instruction to the hardware module, the second instruction being used to instruct the processor to utilize its processing resources. The algorithm module receives processing resources from the processor, inputs the road image to be identified into the trained image recognition model, and calls the processor's processing resources to generate a target road image. The target road image is the road image to be identified marked with a target identifier, which is used to indicate the lane lines. The training process of the image recognition model includes: training a floating-point first image recognition model using multiple preset road images marked with lane line indicators; converting the floating-point first image recognition model into an integer second image recognition model; performing quantization perception training on the second image recognition model based on the multiple preset road images to obtain the trained image recognition model; converting the trained image recognition model into a DPU file; and deploying the DPU file in the processor to improve the compatibility between the image recognition model and the processor in the target device and reduce the image recognition time of the target device. The algorithm module sends the target road image to the transmission module; The transmission module receives the target road image from the algorithm module and determines whether the image parameters of the target road image are greater than the preset parameter threshold. The image parameters include at least one of the following: resolution, saturation, and image size. If the image parameters of the target road image are greater than the preset parameter threshold, the target road image is sent to the display module through the transmission module. The display module receives the target road image from the transmission module and displays the target road image. If the image parameters of the target road image are less than or equal to the preset parameter threshold, an alarm message is sent to the service module through the transmission module. The alarm message is used to indicate an image recognition anomaly. In response to the alarm message, the service module sends the first instruction to the algorithm module.

2. The method according to claim 1, characterized in that, The process of converting the first image recognition model (floating-point type) into the second image recognition model (integer type) includes: Based on preset weights, the floating-point type first image recognition model is converted into an integer type second image recognition model.

3. A lane line image recognition device, characterized in that, Applied to a target device, the target device is equipped with a processor and an image recognition model. The target device includes: an acquisition module, a hardware module, an algorithm module, a service module, a transmission module, and a display module. The hardware module is equipped with the processor, and the algorithm module is equipped with a trained image recognition model. The target device stores preset parameter thresholds. The device includes: The acquisition unit is used to acquire a road image to be identified through the acquisition module, wherein the road image to be identified includes lane lines; The processing unit is configured to: send a first instruction to the algorithm module via the service module, the first instruction instructing the acquisition of the road image to be identified from the acquisition module and input the road image to be identified into the trained image recognition model; in response to the first instruction, the algorithm module acquires the road image to be identified from the acquisition module and sends a second instruction to the hardware module, the second instruction instructing the mobilization of the processor's processing resources; receive the processing resources from the processor via the algorithm module, input the road image to be identified into the trained image recognition model, and mobilize the processor's processing resources to generate a target road image, the target road image being marked with a target identifier. The road image to be identified, wherein the target identifier is used to indicate the lane lines; the training process of the image recognition model includes: training a floating-point first image recognition model using multiple preset road images marked with lane line indicator identifiers; converting the floating-point first image recognition model into an integer second image recognition model; performing quantization perception training on the second image recognition model based on the multiple preset road images to obtain the trained image recognition model; converting the trained image recognition model into a DPU file; and deploying the DPU file in the processor to improve the compatibility between the image recognition model and the processor in the target device, reduce the image recognition time of the target device, and improve recognition efficiency; The processing unit is further configured to determine whether the image parameters of the target road image are greater than the preset parameter threshold, wherein the image parameters include at least one of the following: resolution, saturation, and image size; if the image parameters of the target road image are less than or equal to the preset parameter threshold, an alarm message indicating an image recognition anomaly is output and a re-recognition is triggered. The display unit is used to display the target road image when the image parameters of the target road image are greater than the preset parameter threshold.

4. The apparatus according to claim 3, characterized in that, The processing unit is further configured to convert the floating-point type first image recognition model into an integer type second image recognition model according to a preset weight.

5. An electronic device, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method as described in any one of claims 1 to 2.

6. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is capable of performing the method as described in any one of claims 1 to 2.

Citation Information

Patent Citations

  • Lane line recognition method, device and system and vehicle

    CN111611830A

  • Image recognition method and device, electronic equipment and storage medium

    CN114463580A