Lane line detection method and device based on artificial intelligence

Abstract

The invention discloses a lane line detection method and device based on artificial intelligence. The method comprises the following steps: acquiring image information of a road to be detected; preprocessing the to-be-detected road image information to obtain target processing path information; and performing detection processing on the target processing path information to obtain target lane line detection result information.

Description

Technical Field

[0001] The present invention relates to the field of detection technologies, and in particular, to a lane line detection method and device based on artificial intelligence. Background Art

[0002] Traditional lane line detection methods face various challenges, including light changes, diversity of road materials and colors, occlusion problems, road wear and aging, complex road geometries, road condition changes, nearly parallel non-lane lines, viewpoints and camera calibration, and real-time requirements. Traditional methods require manual adjustment of operators and thresholds, resulting in a large workload and poor robustness, and the detection results are not ideal in complex environments. Therefore, a lane line detection method and device based on artificial intelligence are provided to improve the efficiency and accuracy of lane line detection. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a lane line detection method and device based on artificial intelligence, which are beneficial to improving the efficiency and accuracy of lane line detection.

[0004] To solve the above technical problem, in the first aspect of an embodiment of the present invention, a lane line detection method based on artificial intelligence is disclosed, and the method includes:

[0005] Obtaining information of a road image to be detected;

[0006] Performing preprocessing on the information of the road image to be detected to obtain target processed path information;

[0007] Performing detection processing on the target processed path information to obtain target lane line detection result information.

[0008] In the second aspect of an embodiment of the present invention, a lane line detection device based on artificial intelligence is disclosed, and the device includes:

[0009] An obtaining module, configured to obtain information of a road image to be detected;

[0010] A first processing module, configured to perform preprocessing on the information of the road image to be detected to obtain target processed path information;

[0011] A second processing module, configured to perform detection processing on the target processed path information to obtain target lane line detection result information.

[0012] In the third aspect of the present invention, another lane line detection device based on artificial intelligence is disclosed, and the device includes:

[0013] A memory storing executable program code;

[0014] A processor coupled to the memory;

[0015] The processor calls the executable program code stored in the memory and executes some or all of the steps in the lane line detection method based on artificial intelligence disclosed in the first aspect of the embodiments of the present invention.

[0016] The fourth aspect of the present invention discloses a computer-readable storage medium storing computer instructions that, when called, are used to execute some or all of the steps in the lane line detection method based on artificial intelligence disclosed in the first aspect of the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0018] Figure 1 is a schematic diagram of the scenario of the lane line detection system based on artificial intelligence provided by the embodiments of the present invention;

[0019] Figure 2 is a schematic flowchart of a lane line detection method based on artificial intelligence disclosed in the embodiments of the present invention;

[0020] Figure 3 is a schematic structural diagram of a lane line detection device based on artificial intelligence disclosed in the embodiments of the present invention;

[0021] Figure 4 is a schematic structural diagram of another lane line detection device based on artificial intelligence disclosed in the embodiments of the present invention;

[0022] Figure 5 is a schematic structural diagram of a target feature extraction model disclosed in the embodiments of the present invention;

[0023] Figure 6 is a schematic structural diagram of a fifth extraction module disclosed in the embodiments of the present invention;

[0024] Figure 7 is a schematic structural diagram of a target detection model disclosed in the embodiments of the present invention;

[0025] Figure 8 is a schematic flowchart of another lane line detection method based on artificial intelligence disclosed in the embodiments of the present invention;

[0026] Figure 9 is a schematic diagram of the overall structure of the CondLaneNet network disclosed in the embodiments of the present invention. Detailed implementation manners

[0027] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having", and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0029] Referring to "embodiment" herein means that a specific feature, structure or characteristic described in connection with the embodiment can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0030] In this application, the term "exemplary" is used to mean "serving as an example, illustration, or description". Any embodiment described as "exemplary" in this application is not necessarily to be construed as more preferred or more advantageous than other embodiments. In order to enable any person skilled in the art to implement and use this application, the following description is given. In the following description, details are set forth for the purpose of explanation. It should be understood that those of ordinary skill in the art can recognize that this application can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed in this application.

[0031] It should be noted that since the method of the embodiment of this application is executed in a computer device, the processing objects of each computer device exist in the form of data or information. For example, time, which is actually time information. It can be understood that in subsequent embodiments, if dimensions, quantities, positions, etc. are mentioned, they are all corresponding data existences for the computer device to process, and specific details are not elaborated here.

[0032] It should be noted that a brief introduction to the artificial intelligence-related technologies that may be involved in this application is provided. Artificial Intelligence (AI) uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, including theories, methods, technologies, and application systems for perceiving the environment, acquiring knowledge, and using knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable the machines to have the functions of perception, reasoning, and decision-making.

[0033] Artificial intelligence technology is an interdisciplinary subject with a wide range of fields involved, including both hardware-level and software-level technologies. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technologies mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0034] Computer Vision Technology (CV) Computer vision is a science that studies how to enable machines to "see". More specifically, it refers to using cameras and computers to replace human eyes for tasks such as target recognition and measurement in machine vision, and further performing image processing to make the images processed by the computer more suitable for human eyes to observe or be transmitted to instruments for detection. As a scientific discipline, computer vision studies related theories and technologies and attempts to establish artificial intelligence systems that can obtain information from images or multi-dimensional data. Computer vision technology usually includes technologies such as image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, and also includes common biometric recognition technologies such as face recognition and fingerprint recognition.

[0035] Unimodal information is data of only one type, such as one of the data information types like text, image, audio, video, electromagnetic signals, etc. Multimodal information is data information that includes at least two types of unimodal information. Further, multimodal information is applicable to complex tasks that require integrating multiple information sources, such as sentiment analysis, robot interaction, and autonomous driving. By integrating information of multiple modalities, higher performance and accuracy can usually be achieved in tasks.

[0036] A large model refers to an artificial neural network model with a very large number of parameters. In the field of artificial intelligence, a large model usually refers to a model with hundreds of millions to trillions of parameters. The model usually needs to be trained on a large-scale dataset and requires a large amount of computing resources for optimization and adjustment. Large models are usually used to solve complex tasks such as natural language processing, computer vision, and speech recognition. Generative AI is a type of AI that can create new content and ideas, including conversations, stories, images, videos, and music. In the embodiments of this application, the large model can be large-scale language models such as ChatGPT, BERT, XLNet, Zhipu Model, Claude, Moonshot AI Model, ChatGLM Model, Tongwen Qianyi Model, MiniMax Model, Spark Model, Llama Model, 360GPT Model, Qwen Model, Baichuan Model, Lark Model, vivoLM Model, and Wenxin Yiyan, and the embodiments of this application do not make any limitations.

[0037] The embodiments of this application provide an artificial intelligence-based lane line detection method, device, computer device, and computer-readable storage medium, which will be described in detail below.

[0038] Please refer to Figure 1 , Figure 1 which is a schematic diagram of the scenario of the artificial intelligence-based lane line detection system provided by the embodiments of this application. The artificial intelligence-based lane line detection system may include a computer device 100, and the computer device 100 integrates an artificial intelligence-based lane line detection device, such as Figure 1 the computer device in

[0039] In the embodiments of this application, the computer device 100 is mainly used to obtain the information of the road image to be detected;

[0040] preprocess the information of the road image to be detected to obtain the target processed path information;

[0041] perform detection processing on the target processed path information to obtain the target lane line detection result information.

[0042] It can improve the efficiency and accuracy of lane line detection.

[0043] In the embodiments of this application, the computer device 100 can be an independent server or a server network or server cluster composed of servers. For example, the computer device 100 described in the embodiments of this application includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud server composed of multiple servers. Among them, the cloud server is composed of a large number of computers or network servers based on cloud computing.

[0044] It can be understood that the computer device 100 used in the embodiments of the present application can be a device that includes both receiving and transmitting hardware, that is, a device with receiving and transmitting hardware capable of performing two-way communication on a two-way communication link. Such a device may include: cellular or other communication devices, which have a single-line display or a multi-line display or cellular or other communication devices without a multi-line display. Specifically, the computer device 100 can be a desktop terminal or a mobile terminal, and the computer device 100 can also be a mobile phone, a tablet computer, a laptop computer, etc.

[0045] Those skilled in the art can understand that Figure 1 the application environment shown in Figure 1 is only one application scenario of the solution of the present application, and does not constitute a limitation on the application scenario of the solution of the present application. Other application environments may also include more or fewer computer devices than those shown in Figure 1 For example, only 1 computer device is shown in

[0046] In addition, as shown in Figure 1 the lane line detection system based on artificial intelligence may also include a memory 200 for storing data, such as image data, location information, etc.

[0047] It should be noted that Figure 1 the scenario schematic diagram of the lane line detection system based on artificial intelligence shown in

[0048] The present invention discloses a lane line detection method and device based on artificial intelligence, which are beneficial to improving the efficiency and accuracy of lane line detection. The following will be described in detail respectively.

[0049] Embodiment 1

[0050] Please refer to Figure 2 Figure 2 which is a schematic flowchart of a lane line detection method based on artificial intelligence disclosed in the embodiments of the present invention. Among them, Figure 2 the described lane line detection method based on artificial intelligence is applied to a management system, such as a local server or a cloud server for management, etc., which is not limited in the embodiments of the present invention. As shown in Figure 2 ​As shown, the artificial intelligence-based lane line detection method may include the following operations:

[0051] 101. Obtain the road image information to be detected.

[0052] 102. Preprocess the road image information to be detected to obtain the target processing path information.

[0053] 103. Perform detection processing on the target processing path information to obtain the target lane line detection result information.

[0054] It should be noted that the above road image information to be detected may be based on the images collected by a camera device. For example, the images collected by a camera installed on the top of a vehicle can effectively capture the road conditions without affecting the driver's line of sight. The embodiments of the present invention are not limited thereto.

[0055] It should be noted that the system of the artificial intelligence-based lane line detection method based on the present application is an end-to-end processing method with less post-processing and high real-time performance. The embodiments of the present invention are not limited thereto.

[0056] It can be seen that implementing the artificial intelligence-based lane line detection method described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0057] In an optional embodiment, performing detection processing on the target processing path information to obtain the target lane line detection result information includes:

[0058] Performing feature extraction processing on the target processing path information by using a target feature extraction model to obtain target road feature information;

[0059] Performing detection processing on the target road feature information by using a target detection model to obtain the target lane line detection result information.

[0060] It should be noted that the above target detection model uses the extracted road surface and lane marking features to generate a lane line feature representation and infer the lane line position, which can improve the detection accuracy and effectively improve the speed and accuracy of lane line detection. The embodiments of the present invention are not limited thereto.

[0061] It should be noted that the above target feature extraction model extracts features from the preprocessed road image to extract rich features such as color, edge, texture, and other specific lane line image information, laying a foundation for subsequent lane line detection. The embodiments of the present invention are not limited thereto.

[0062] It should be noted that the above-mentioned target road feature information includes first target road feature information and second target road feature information, which are not limited in the embodiments of the present invention. The feature scales of the first target road feature information and the second target road feature information are inconsistent, so as to achieve accurate detection of information in different dimensions of the subsequent lane lines, which is not limited in the embodiments of the present invention.

[0063] It should be noted that the above-mentioned target lane line detection result information includes first target detection result information and second target detection result information, which are not limited in the embodiments of the present invention. Further, the above-mentioned first target detection result information represents the pixel point information along the extending direction of the lane line (such as the range of the currently detected lane line in the vehicle's forward direction), and the second target detection result information represents the information indicating the width across the lane line (such as the range of the currently detected lane line deviating from the forward direction), which are not limited in the embodiments of the present invention.

[0064] It can be seen that implementing the lane line detection method based on artificial intelligence described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0065] In another optional embodiment, as Figure 5 shown, the target feature extraction model includes a first convolution module 401, a first pooling module 402, a first extraction module, a second extraction module 404, a third extraction module 405, a fourth extraction module 406, and a fifth extraction module 407; wherein,

[0066] The input end of the first convolution module is configured to receive the first model input of the target feature extraction model, and the output end of the first convolution module is connected to the input end of the first pooling module; the output end of the first pooling module is connected to the input end of the first extraction module; the output end of the first extraction module is respectively connected to the input ends of the second extraction module and the fifth extraction module; the output end of the second extraction module is respectively connected to the input ends of the third extraction module and the fifth extraction module; the output end of the third extraction module is respectively connected to the input ends of the fourth extraction module and the fifth extraction module; the output end of the fourth extraction module is connected to the input end of the fifth extraction module; the output end of the fifth extraction module is configured to output the first model output and the second model output of the target feature extraction model.

[0067] It should be noted that the convolution kernel size of the above-mentioned first convolution module is 7×7, the number of convolution kernels is 64, and the number of channels is 1, which are not limited in the embodiments of the present invention.

[0068] It should be noted that the above-mentioned first pooling module is constructed based on a max pooling layer, which is not limited in the embodiments of the present invention.

[0069] It should be noted that the model architectures of the above-mentioned first extraction module, second extraction module, third extraction module, and fourth extraction module are the same, and are not limited in the embodiments of the present invention.

[0070] It should be noted that the above-mentioned first model input is target processing path information, and is not limited in the embodiments of the present invention.

[0071] It should be noted that the above-mentioned first model output and second model output are the first target detection result information and the second target detection result information respectively, and are not limited in the embodiments of the present invention.

[0072] It should be noted that the training loss function of the above-mentioned target feature extraction model can be a cross-entropy loss function, and is not limited in the embodiments of the present invention.

[0073] It should be noted that through the layer-by-layer feature extraction of the convolutional module, pooling module, and multiple extraction modules in the above-mentioned target feature extraction model, the extraction of gradually advanced road features can be realized, the enhanced extraction and output of global and local features can be achieved, so as to fully extract the high-resolution image information and low-resolution information in the image, which is more conducive to the accurate lane line detection of the target detection model, and is not limited in the embodiments of the present invention.

[0074] It can be seen that implementing the lane line detection method based on artificial intelligence described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0075] In another optional embodiment, as Figure 5 shown, the first extraction module includes a first convolutional unit, a second convolutional unit, a third convolutional unit, a fourth convolutional unit, a fifth convolutional unit, a sixth convolutional unit, a seventh convolutional unit, an eighth convolutional unit, a ninth convolutional unit, a first activation unit, a second activation unit, a third activation unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a first normalization unit, a second normalization unit, a third normalization unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a first fusion unit, and a second fusion unit; wherein,

[0076] The input end of the first convolutional unit is configured as the input end of the first extraction module. The output end of the first convolutional unit is respectively connected to the input end of the second convolutional unit, the input end of the fifth convolutional unit, and the input end of the first fusion unit. The output end of the second convolutional unit is connected to the input end of the first normalization unit. The output end of the first normalization unit is connected to the input end of the first activation unit. The output end of the first activation unit is connected to the input end of the third convolutional unit. The output end of the third convolutional unit is connected to the input end of the second normalization unit. The output end of the second normalization unit is connected to the input end of the second activation unit. The output end of the second activation unit is connected to the input end of the fourth convolutional unit. The output end of the fourth convolutional unit is connected to the input end of the third normalization unit. The output end of the third normalization unit is connected to the input end of the first fusion unit. The output end of the fifth convolutional unit is connected to the input end of the fourth normalization unit. The output end of the fourth normalization unit is connected to the input end of the first fusion unit. The output end of the first fusion unit is connected to the input end of the third activation unit. The output end of the third activation unit is connected to the input end of the sixth convolutional unit, the input end of the ninth convolutional unit, and the input end of the second fusion unit. The output end of the sixth convolutional unit is connected to the input end of the fifth normalization unit. The output end of the fifth normalization unit is connected to the input end of the fourth activation unit. The output end of the fourth activation unit is connected to the input end of the seventh convolutional unit. The output end of the seventh convolutional unit is connected to the input end of the sixth normalization unit. The output end of the sixth normalization unit is connected to the input end of the fifth activation unit. The output end of the fifth activation unit is connected to the input end of the eighth convolutional unit. The output end of the eighth convolutional unit is connected to the input end of the seventh normalization unit. The output end of the seventh normalization unit is connected to the output end of the second fusion unit. The output end of the ninth convolutional unit is connected to the input end of the eighth normalization unit. The output end of the eighth normalization unit is connected to the input end of the second fusion unit. The output end of the second fusion unit is connected to the input end of the sixth activation unit. The output end of the sixth activation unit is configured as the output end of the first extraction module.

[0077] It should be noted that the above first extraction model can extract information on shallow details of roads such as brightness and edges through the layer-by-layer deep stacking of convolutional units and the information fusion of multiple cross-layer fusion units, while retaining the information of the original road image features (for the subsequent deep feature extraction of the first to fourth extraction modules by the fifth extraction module), so as to achieve the extraction of rich semantic road feature information. The embodiments of the present invention are not limited thereto.

[0078] It should be noted that the above first fusion unit and second fusion unit are constructed based on element-wise addition operations. The embodiments of the present invention are not limited thereto.

[0079] It should be noted that the above first activation unit, second activation unit, third activation unit, fourth activation unit, fifth activation unit, and sixth activation unit are all constructed based on the ReLU activation function. The embodiments of the present invention are not limited thereto.

[0080] It should be noted that the above first normalization unit, second normalization unit, third normalization unit, fourth normalization unit, fifth normalization unit, sixth normalization unit, seventh normalization unit, and eighth normalization unit are constructed based on the batch normalization layer, which is not limited in the embodiments of the present invention.

[0081] It should be noted that the convolution kernel size of the above first convolution unit is 7×7, the number of convolution kernels is 64, and the number of channels is 1, which is not limited in the embodiments of the present invention. The convolution kernel sizes of the above second convolution unit, fourth convolution unit, sixth convolution unit, and eighth convolution unit are 1×1, the number of convolution kernels is 64, and the number of channels is 1, which is not limited in the embodiments of the present invention. The convolution kernel sizes of the above third convolution unit, fifth convolution unit, seventh convolution unit, and ninth convolution unit are 3×3, the number of convolution kernels is 256, and the number of channels is 1, which is not limited in the embodiments of the present invention.

[0082] It can be seen that implementing the artificial intelligence-based lane line detection method described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0083] In another alternative embodiment, as Figure 6 shown, the fifth extraction module includes a tenth convolution unit 4061, an eleventh convolution unit 4062, a twelfth convolution unit 4063, a thirteenth convolution unit 4064, a fourteenth convolution unit 4067, a fifteenth convolution unit 40610, a sixteenth convolution unit 40612, a seventeenth convolution unit 40613, an eighteenth convolution unit 40620, a third fusion unit 4065, a fourth fusion unit 4068, a fifth fusion unit 40611, a sixth fusion unit 40619, a first sampling unit 4066, a second sampling unit 4069, a third sampling unit 40623, a seventh activation unit 40615, an eighth activation unit 40622, a ninth normalization unit 40614, a tenth normalization unit 40621, a first slicing unit 40617, a dimension conversion unit 40618, and a multi-head attention unit 40616; wherein,

[0084] The input end of the tenth convolutional unit is connected to the output end of the first extraction module, and the output end of the tenth convolutional unit is connected to the input end of the fifth fusion unit; the input end of the eleventh convolutional unit is connected to the output end of the second extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the thirteenth convolutional unit; the output end of the thirteenth convolutional unit is connected to the input end of the third fusion unit; the input end of the twelfth convolutional unit is connected to the output end of the third extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the fourteenth convolutional unit; the output end of the fourteenth convolutional unit is connected to the input end of the fourth fusion unit; the input end of the seventeenth convolutional unit is connected to the output end of the fourth extraction module, and the output end of the seventeenth convolutional unit is connected to the input end of the ninth normalization unit; the output end of the ninth normalization unit is connected to the input end of the seventh activation unit; the output end of the seventh activation unit is respectively connected to the input ends of the multi-head attention unit and the first slicing unit; the output end of the multi-head attention unit is connected to the input end of the sixth fusion unit; the output end of the first slicing unit is connected to the input end of the dimension transformation unit; the output end of the dimension transformation unit is connected to the input end of the sixth fusion unit; the output end of the sixth fusion unit is connected to the input end of the eighteenth convolutional unit; the output end of the eighteenth convolutional unit is connected to the input end of the tenth normalization unit; the output end of the tenth normalization unit is connected to the input end of the eighth activation unit; the output end of the eighth activation unit is connected to the input end of the third sampling unit; the output end of the third sampling unit is connected to the input end of the fourth fusion unit; the output end of the fourth fusion unit is respectively connected to the input ends of the second sampling unit and the fifteenth convolutional unit; the output end of the second sampling is connected to the input end of the third fusion unit; the output end of the third fusion unit is connected to the input end of the first sampling unit; the output end of the first sampling unit is connected to the input end of the fifth fusion unit; the output end of the fifth fusion unit is connected to the input end of the sixteenth convolutional unit; the output end of the sixteenth convolutional unit is configured to output the first model output of the target feature extraction model; the output end of the fifteenth convolutional unit is configured to output the second model output of the target feature extraction model.

[0085] It should be noted that the above-mentioned fifth extraction module realizes the in-depth extraction of feature information through convolution and sampling operations on the features of different scales extracted by the first to fourth extraction modules, and then performs the fusion process of element-wise addition, so as to realize the effective fusion of features of different depths and high and low resolution image features, and realize the extraction of rich semantic feature information, which is more conducive to the detection and recognition processing of the target detection model. The embodiments of the present invention do not make any limitations. Further, after the above-mentioned conversion processing of the data dimension after the feature information obtained by the fourth extraction module is based on the multi-head attention mechanism and convolution, and then the two kinds of data are fused, the extraction of global features can be further strengthened, which is more conducive to the accurate detection, recognition and prediction analysis of lane lines. The embodiments of the present invention do not make any limitations.

[0086] It should be noted that the above first sampling unit, second sampling unit, and third sampling unit are constructed based on the upsampling operation, and the embodiments of the present invention do not make any limitations.

[0087] It should be noted that the above third fusion unit, fourth fusion unit, fifth fusion unit, and sixth fusion unit are constructed based on the element-wise addition operation, and the embodiments of the present invention do not make any limitations.

[0088] It should be noted that the above seventh activation unit and eighth activation unit are constructed based on the RELU activation function, and the embodiments of the present invention do not make any limitations.

[0089] It should be noted that the above ninth normalization unit and tenth normalization unit are constructed based on the batch normalization layer, and the embodiments of the present invention do not make any limitations.

[0090] It should be noted that the above first slicing unit is constructed based on the FLAATEN operation. First, the data is converted into one-dimensional data so that the subsequent dimension conversion unit can reconstruct the data dimension. The embodiments of the present invention do not make any limitations.

[0091] It should be noted that the above multi-head attention unit is constructed based on the multi-head attention mechanism, and the embodiments of the present invention do not make any limitations.

[0092] It should be noted that the above dimension conversion unit is constructed based on the reshape function to construct the data into the same dimension as the output data of the multi-head attention unit. The embodiments of the present invention do not make any limitations.

[0093] It should be noted that the convolution kernels of the above tenth convolution unit, eleventh convolution unit, twelfth convolution unit, thirteenth convolution unit, and fourteenth convolution unit are 1×1, and the number of channels is 1. The embodiments of the present invention do not make any limitations.

[0094] It should be noted that the convolution kernels of the above seventeenth convolution unit and eighteenth convolution unit, seventeenth convolution unit and eighteenth convolution unit are 3×3, and the number of channels is 1. The embodiments of the present invention do not make any limitations.

[0095] It can be seen that implementing the lane line detection method based on artificial intelligence described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0096] In an optional embodiment, the above such as Figure 7As shown in the figure, the target detection model includes a second convolutional module 501, a third convolutional module 502, a fourth convolutional module 507, a fifth convolutional module 508, a sixth convolutional module 511, a seventh convolutional module 512, an eighth convolutional module 513, a ninth convolutional module 516, a tenth convolutional module 519, a first activation module 505, a second activation module 506, a third activation module 515, a fourth activation module 518, a second pooling module 503, a third pooling module 504, a fourth pooling module 514, a fifth pooling module 517, a first network module 509, and a second network module 510; among them,

[0097] The input ends of the second convolutional module and the third convolutional module are configured to receive the output of the first model; the output end of the second convolutional module is connected to the input end of the second pooling module; the output end of the second pooling module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the fourth convolutional module; the output end of the fourth convolutional module is connected to the input end of the first network module; the output end of the first network module is respectively connected to the input end of the second network module and the input end of the sixth convolutional module; the output end of the third convolutional module is connected to the input end of the third pooling module; the output end of the third pooling module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fifth convolutional module; the output end of the fifth convolutional module is connected to the input end of the second network module; the output end of the second network module is connected to the input end of the seventh convolutional module; the input end of the eighth convolutional module is configured to receive the output of the second model; the output end of the eighth convolutional module is connected to the input end of the fourth pooling module; the output end of the fourth pooling module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the ninth convolutional module; the output end of the ninth convolutional module is connected to the input end of the fifth pooling module; the output end of the fifth pooling module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the tenth convolutional module; the output end of the tenth convolutional module is respectively connected to the input end of the sixth convolutional module and the input end of the seventh convolutional module; the output end of the sixth convolutional module is configured to output the third model output of the target detection model; the output end of the seventh convolutional module is configured to output the fourth model output of the target detection model.

[0098] It should be noted that the target detection model performs multiple convolutional operations on different-scale feature information respectively, and then divides into two branches. After the associated feature information based on the long short-term memory network, the dynamic convolutional layer respectively detects the lane line features, so as to obtain the lane line pixel information in two-dimensional directions representing the longitudinal and width directions of the lane line, in order to achieve accurate detection of the lane line. The embodiments of the present invention are not limited thereto.

[0099] It should be noted that the above first activation module, second activation module, third activation module, and fourth activation module are constructed based on the ReLU activation function, and the embodiments of the present invention are not limited thereto.

[0100] It should be noted that the above second pooling module, third pooling module, fourth pooling module, and fifth pooling module are constructed based on the max pooling layer, and the embodiments of the present invention are not limited thereto.

[0101] It should be noted that the above first network module and second network module are based on the long short-term memory network architecture to strengthen the analysis of the correlation between different scale feature information, and the embodiments of the present invention are not limited thereto.

[0102] It should be noted that the above sixth convolution module and seventh convolution module are constructed based on the dynamic convolution layer to dynamically adapt to the data size after convolution, normalization, and activation operations on the data from the outputs of the first model and the second model, and then perform classification recognition in the same convolution layer. The embodiments of the present invention are not limited thereto.

[0103] It should be noted that the convolution kernels of the above second convolution module, third convolution module, fourth convolution module, fifth convolution module, eighth convolution module, ninth convolution module, and tenth convolution module have a size of 3×3 and a channel number of 1. The embodiments of the present invention are not limited thereto.

[0104] It should be noted that the training loss function of the above object detection model can be the cross-entropy loss function, and the embodiments of the present invention are not limited thereto.

[0105] It can be seen that implementing the lane line detection method based on artificial intelligence described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0106] In another optional embodiment, preprocessing the road image information to be detected to obtain target processed path information, including:

[0107] Performing noise reduction and enhancement processing on the road image information to be detected to obtain the first processed road information;

[0108] Performing color space conversion processing on the first processed road information to obtain the second processed road information;

[0109] Performing image cropping processing on the second processed road information to obtain the target processed road information.

[0110] It should be noted that the above noise reduction and enhancement processing of the road image information to be detected is to first perform filtering processing and then perform image enhancement to highlight the lane line features, which is not limited in the embodiments of the present invention. Further, the above image enhancement can be implemented by means such as histogram equalization, piecewise linear stretching function, TopHat algorithm, etc., which is not limited in the embodiments of the present invention.

[0111] In this alternative embodiment, as an alternative implementation manner, the above noise reduction and enhancement processing of the road image information to be detected to obtain the first processed road information includes:

[0112] Performing low-pass filtering processing on the road image information to be detected to obtain the first processed image information;

[0113] Performing median filtering processing on the first processed image information to obtain the second processed image information;

[0114] Performing quadratic linear interpolation correction processing on the third processed image to obtain the third processed image;

[0115] Performing image enhancement processing on the second processed image information to obtain the first processed road information.

[0116] It should be noted that the above low-pass filtering and median filtering processing of the image can further reduce the influence of different burrs on the detailed features such as the lane line edge. Performing quadratic linear interpolation correction processing on the third processed image to further reduce image noise and improve image quality, making the lane lines in the image clearer and easier to detect and recognize, which is not limited in the embodiments of the present invention.

[0117] It should be noted that the above color space conversion processing of the first processed road information is to convert the image from the RGB space to a color space (such as Lab, hue-brightness-saturation, etc.) to enhance the saliency analysis of the lane lines, which is not limited in the embodiments of the present invention.

[0118] It should be noted that the above image cropping processing of the second processed road information is to select the ROI area to remove the redundant content in the image and determine the bottom of the lane image, thereby improving the lane line detection speed and accuracy, which is not limited in the embodiments of the present invention.

[0119] It can be seen that implementing the lane line detection method based on artificial intelligence described in the embodiments of the present invention is beneficial to improving the efficiency and accuracy of lane line detection.

[0120] Embodiment 2

[0121] Please refer to Figure 8 , Figure 8 which is a schematic flowchart of another lane line detection method based on artificial intelligence disclosed in the embodiments of the present invention. Among them,Figure 8 The described artificial intelligence-based lane line detection method can be applied to the field of detection technology, which is not limited in the embodiments of the present invention. As Figure 8 shown, the artificial intelligence-based lane line detection method may include:

[0122] 1) Extract images from the video captured by the camera and use them as the input of the lane line detection system. The camera for collecting images is generally installed on the top of the vehicle, which can effectively capture road conditions without affecting the driver's line of sight.

[0123] 2) In the image preprocessing stage, perform denoising, enhancement, color space conversion, and region of interest (ROI) selection on the collected video images. First, perform linear filtering or non-linear filtering on the collected video images; secondly, use histogram equalization, piecewise linear stretching function, and TopHat algorithm to enhance the features of the lane lines; in order to highlight the features of the lane lines, convert from the RGB space to the Lab, Hue-Intensity-Saturation (HIS) color space for saliency analysis; in order to remove redundant content in the image and improve the detection speed and accuracy, limit the processing area ROI at the bottom of the lane image.

[0124] 3) In the selected ROI, extract color, edge, and texture lane line image information. Based on color, edge, and texture information, extract important features for lane line detection such as the contrast between the lane line and the road surface, the constant width of the lane and the lane line, the curvature of the lane line, and the parallel characteristics of the lane line.

[0125] Perform feature fusion on the contrast between the lane line and the road surface, the constant width of the lane and the lane line, the curvature of the lane line, and the parallel characteristics of the lane line to obtain fused feature parameters;

[0126] The fusion method is:

[0127] Optionally, the two feature matrices to be fused are X and Y respectively. X is an n×m-dimensional matrix, Y is a p×m-dimensional matrix, M represents the number of samples, n and p represent the dimensions of the two features. Project the two matrices onto 1D for linear representation, corresponding to the projection vectors a and b respectively. Then the projected feature matrices become:

[0128] X' = a T X, Y' = b T Y

[0129] The purpose is to maximize the correlation coefficient between X' and Y', so as to obtain the projection vectors a and b when the correlation coefficient is the largest, that is:

[0130]

[0131] Before projection, the data is standardized. The purpose of standardization is to make the mean of the data 0 and the variance 1, and we can get:

[0132] cov(X',Y') = cov(a T X,b T Y) = E(<a T X,b T Y>) = E((a T X)(b T Y) T ) = a T E(XY T )b

[0133] D(X') = D(a T X) = a T E(XX T )a

[0134] D(Y') = D(b T Y) = b T E(YY T )b

[0135] D(X) = cov(X,X) = E(XX T )

[0136] D(Y) = cov(Y,Y) = E(YY T )

[0137] cov(X,Y) = E(XY T ),cov(Y,X) = E(YX T )

[0138] S XX = cov(X,X), then the solution target is transformed into:

[0139]

[0140] Step1: Calculate the variances S XX of X and Y, and S YY , and the covariances S XY = S YX T ;

[0141] Step2: Calculate the matrix

[0142] Step3: Solve the singular values of M to obtain the largest singular value and its front and back singular vectors u, v;

[0143] Step4: The projection vectors a and b of X and Y are respectively: X'=a T X,Y'=b T Y

[0144] The fusion result is Z = λ1X′+λ2Y′, λ1+λ2=1, λ11 and λ2 are weights.

[0145] The contrast between the lane line and the road surface and the constant width of the lane and the lane line are fused to obtain a first fusion feature;

[0146] The lane line curvature and the lane line parallel characteristics are fused to obtain a second fusion feature;

[0147] Fuse the first fusion feature and the second fusion feature to obtain the fusion feature parameter

[0148] 4) In the lane line detection stage, the lane line is detected using feature-based, model-based, learning-based and other methods based on the extracted fusion feature parameters.

[0149] The feature-based method uses the extracted road surface and lane marking features to detect lane markings through segmentation and clustering. The model-based method uses mathematical methods to build a lane line model. When the lane line is broken or partially blocked, the position, angle, curvature and other information provided by the model helps improve the performance of lane line detection.

[0150] The learning-based method automatically generates lane feature representation and infers the lane position by learning a data set containing a large number of samples.

[0151] The available learning-based methods are:

[0152] The detection model is trained using the fusion feature parameters to obtain an optimized detection model;

[0153] Using the optimized detection model, the fused features to be processed are detected to obtain the lane line detection results;

[0154] The detection model is: CondLaneNet network. CondLaneNet combines the advantages of instance segmentation and lane classification to solve the lane instance level recognition problem. The optimization focus of CondLaneNet is the lane line shape specified by the lane direction formula.

[0155] The CondLaneNet network divides lane line detection into two steps: instance detection and shape prediction. Instance detection can predict each lane line instance and regress a set of dynamic convolution kernel parameters for each instance (lane line). Shape prediction uses conditional convolution to refine the shape of the lane line. Since each lane line instance corresponds to a set of convolution parameters, the lane line prediction is at the instance level.

[0156] The overall structure of the CondLaneNet network is as Figure 9 shown, and it consists of three parts: Backbone, Proposalhead, and Condition shape head. The Backbone is responsible for extracting image features; the Proposal head is used to detect lane line instances and generate dynamic convolution kernel parameters for each instance; the Conditional shape head uses the dynamic convolution kernel parameters and conditional convolution generated in the Proposal head step to determine the point set describing the lane line.

[0157] 5) The tracking module realizes lane line prediction to improve the accuracy of detection. It is used to follow lane changes, utilize the lane line position obtained at the previous moment, and estimate the lane line position at the next moment, which can effectively improve the speed and accuracy of lane line detection.

[0158] The tracking method process is as follows:

[0159] Input: Lane line image sequence (for each frame of the image sequence, the candidate regions containing the target are constructed into a graph structure); the position and size of the target in the first frame;

[0160] Initialize the regression label and the transformed graph Laplacian matrix

[0161] for each frame of the image

[0162] if frame>1

[0163] Extract the multi-channel feature X from the candidate regions

[0164] Calculate the K-order filtering response The formula is:

[0165]

[0166] The filtering basis can be expressed as The Laplacian matrix of the lane line image sequence is L n , I is the identity matrix, and the transformation gives λ max is the largest eigenvalue of L n , N is the number of vertices of the graph structure, and Kd is

[0167] Calculate the kernel matrix K, the formula is:

[0168]

[0169] φ(x) high-dimensional space,

[0170] Calculate the confidence score y

[0171]

[0172] α is the eigenvector of y;

[0173] Find the position where the confidence score is the largest and update the position of the tracking target in the current frame

[0174] end

[0175] Extract the multi-channel feature X from the current target position

[0176] Calculate the K-order filtering response

[0177] Calculate the kernel matrix K

[0178] Calculate the kernel regression model based on graph signals

[0179] Update the model

[0180] Output: The position coordinates of the tracking target

[0181] Embodiment III

[0182] Please refer to Figure 3 , Figure 3 is a schematic structural diagram of a lane line detection device based on artificial intelligence disclosed in an embodiment of the present invention. Among them, Figure 3 The described device can be applied to a management system, such as a local server or a cloud server for management, etc., which is not limited in the embodiments of the present invention. As Figure 3 shown, the device may include:

[0183] An acquisition module 201, configured to acquire information of a road image to be detected;

[0184] A first processing module 202, configured to preprocess the information of the road image to be detected to obtain target processing path information;

[0185] A second processing module 203, configured to perform detection processing on the target processing path information to obtain target lane line detection result information.

[0186] It can be seen that implementing Figure 3 The described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0187] In another alternative embodiment, as Figure 3 shown, performing detection processing on the target processing path information to obtain target lane line detection result information includes:

[0188] The target feature extraction model is used to perform feature extraction processing on the target processing path information to obtain target road feature information;

[0189] The target detection model is used to perform detection processing on the target road feature information to obtain target lane line detection result information.

[0190] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0191] In another alternative embodiment, as Figure 3 shown, the target feature extraction model includes a first convolution module, a first pooling module, a first extraction module, a second extraction module, a third extraction module, a fourth extraction module, and a fifth extraction module; among them,

[0192] The input end of the first convolution module is configured to receive the first model input of the target feature extraction model, and the output end of the first convolution module is connected to the input end of the first pooling module; the output end of the first pooling module is connected to the input end of the first extraction module; the output end of the first extraction module is respectively connected to the input ends of the second extraction module and the fifth extraction module; the output end of the second extraction module is respectively connected to the input ends of the third extraction module and the fifth extraction module; the output end of the third extraction module is respectively connected to the input ends of the fourth extraction module and the fifth extraction module; the output end of the fourth extraction module is connected to the input end of the fifth extraction module; the output end of the fifth extraction module is configured to output the first model output and the second model output of the target feature extraction model.

[0193] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0194] In another alternative embodiment, as Figure 3 shown, the first extraction module includes a first convolution unit, a second convolution unit, a third convolution unit, a fourth convolution unit, a fifth convolution unit, a sixth convolution unit, a seventh convolution unit, an eighth convolution unit, a ninth convolution unit, a first activation unit, a second activation unit, a third activation unit, a fourth activation unit, a fifth activation unit, a sixth activation unit, a first normalization unit, a second normalization unit, a third normalization unit, a fourth normalization unit, a fifth normalization unit, a sixth normalization unit, a seventh normalization unit, an eighth normalization unit, a first fusion unit, and a second fusion unit; among them,

[0195] The input end of the first convolutional unit is configured as the input end of the first extraction module. The output end of the first convolutional unit is respectively connected to the input end of the second convolutional unit, the input end of the fifth convolutional unit, and the input end of the first fusion unit. The output end of the second convolutional unit is connected to the input end of the first normalization unit. The output end of the first normalization unit is connected to the input end of the first activation unit. The output end of the first activation unit is connected to the input end of the third convolutional unit. The output end of the third convolutional unit is connected to the input end of the second normalization unit. The output end of the second normalization unit is connected to the input end of the second activation unit. The output end of the second activation unit is connected to the input end of the fourth convolutional unit. The output end of the fourth convolutional unit is connected to the input end of the third normalization unit. The output end of the third normalization unit is connected to the input end of the first fusion unit. The output end of the fifth convolutional unit is connected to the input end of the fourth normalization unit. The output end of the fourth normalization unit is connected to the input end of the first fusion unit. The output end of the first fusion unit is connected to the input end of the third activation unit. The output end of the third activation unit is connected to the input end of the sixth convolutional unit, the input end of the ninth convolutional unit, and the input end of the second fusion unit. The output end of the sixth convolutional unit is connected to the input end of the fifth normalization unit. The output end of the fifth normalization unit is connected to the input end of the fourth activation unit. The output end of the fourth activation unit is connected to the input end of the seventh convolutional unit. The output end of the seventh convolutional unit is connected to the input end of the sixth normalization unit. The output end of the sixth normalization unit is connected to the input end of the fifth activation unit. The output end of the fifth activation unit is connected to the input end of the eighth convolutional unit. The output end of the eighth convolutional unit is connected to the input end of the seventh normalization unit. The output end of the seventh normalization unit is connected to the output end of the second fusion unit. The output end of the ninth convolutional unit is connected to the input end of the eighth normalization unit. The output end of the eighth normalization unit is connected to the input end of the second fusion unit. The output end of the second fusion unit is connected to the input end of the sixth activation unit. The output end of the sixth activation unit is configured as the output end of the first extraction module.

[0196] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is conducive to improving the efficiency and accuracy of lane line detection.

[0197] In another optional embodiment, as Figure 3 shown, the fifth extraction module includes the tenth convolutional unit, the eleventh convolutional unit, the twelfth convolutional unit, the thirteenth convolutional unit, the fourteenth convolutional unit, the fifteenth convolutional unit, the sixteenth convolutional unit, the seventeenth convolutional unit, the eighteenth convolutional unit, the third fusion unit, the fourth fusion unit, the fifth fusion unit, the sixth fusion unit, the first sampling unit, the second sampling unit, the third sampling unit, the seventh activation unit, the eighth activation unit, the ninth normalization unit, the tenth normalization unit, the first slicing unit, the dimension conversion unit, and the multi-head attention unit; where

[0198] The input end of the tenth convolutional unit is connected to the output end of the first extraction module, and the output end of the tenth convolutional unit is connected to the input end of the fifth fusion unit; the input end of the eleventh convolutional unit is connected to the output end of the second extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the thirteenth convolutional unit; the output end of the thirteenth convolutional unit is connected to the input end of the third fusion unit; the input end of the twelfth convolutional unit is connected to the output end of the third extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the fourteenth convolutional unit; the output end of the fourteenth convolutional unit is connected to the input end of the fourth fusion unit; the input end of the seventeenth convolutional unit is connected to the output end of the fourth extraction module, and the output end of the seventeenth convolutional unit is connected to the input end of the ninth normalization unit; the output end of the ninth normalization unit is connected to the input end of the seventh activation unit; the output end of the seventh activation unit is respectively connected to the input ends of the multi-head attention unit and the first slicing unit; the output end of the multi-head attention unit is connected to the input end of the sixth fusion unit; the output end of the first slicing unit is connected to the input end of the dimension conversion unit; the output end of the dimension conversion unit is connected to the input end of the sixth fusion unit; the output end of the sixth fusion unit is connected to the input end of the eighteenth convolutional unit; the output end of the eighteenth convolutional unit is connected to the input end of the tenth normalization unit; the output end of the tenth normalization unit is connected to the input end of the eighth activation unit; the output end of the eighth activation unit is connected to the input end of the third sampling unit; the output end of the third sampling unit is connected to the input end of the fourth fusion unit; the output end of the fourth fusion unit is respectively connected to the input ends of the second sampling unit and the fifteenth convolutional unit; the output of the second sampling is connected to the input end of the third fusion unit; the output end of the third fusion unit is connected to the input end of the first sampling unit; the output end of the first sampling unit is connected to the input end of the fifth fusion unit; the output end of the fifth fusion unit is connected to the input end of the sixteenth convolutional unit; the output end of the sixteenth convolutional unit is configured to output the first model output of the target feature extraction model; the output end of the fifteenth convolutional unit is configured to output the second model output of the target feature extraction model.

[0199] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0200] In yet another alternative embodiment, as Figure 3 shown, the target detection model includes a second convolutional module, a third convolutional module, a fourth convolutional module, a fifth convolutional module, a sixth convolutional module, a seventh convolutional module, an eighth convolutional module, a ninth convolutional module, a tenth convolutional module, a first activation module, a second activation module, a third activation module, a fourth activation module, a second pooling module, a third pooling module, a fourth pooling module, a fifth pooling module, a first network module, and a second network module; wherein,

[0201] The input ends of the second convolutional module and the third convolutional module are configured to receive the output of the first model; the output end of the second convolutional module is connected to the input end of the second pooling module; the output end of the second pooling module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the fourth convolutional module; the output end of the fourth convolutional module is connected to the input end of the first network module; the output end of the first network module is respectively connected to the input end of the second network module and the input end of the sixth convolutional module; the output end of the third convolutional module is connected to the input end of the third pooling module; the output end of the third pooling module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fifth convolutional module; the output end of the fifth convolutional module is connected to the input end of the second network module; the output end of the second network module is connected to the input end of the seventh convolutional module; the input end of the eighth convolutional module is configured to receive the output of the second model; the output end of the eighth convolutional module is connected to the input end of the fourth pooling module; the output end of the fourth pooling module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the ninth convolutional module; the output end of the ninth convolutional module is connected to the input end of the fifth pooling module; the output end of the fifth pooling module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the tenth convolutional module; the output end of the tenth convolutional module is respectively connected to the input end of the sixth convolutional module and the input end of the seventh convolutional module; the output end of the sixth convolutional module is configured to output the third model output of the target detection model; the output end of the seventh convolutional module is configured to output the fourth model output of the target detection model.

[0202] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0203] In another optional embodiment, as Figure 3 shown, preprocess the road image information to be detected to obtain target processing path information, including:

[0204] Perform noise reduction and enhancement processing on the road image information to be detected to obtain the first processed road information;

[0205] Perform color space conversion processing on the first processed road information to obtain the second processed road information;

[0206] Perform image cropping processing on the second processed road information to obtain the target processed road information.

[0207] It can be seen that implementing Figure 3 the described lane line detection device based on artificial intelligence is beneficial to improving the efficiency and accuracy of lane line detection.

[0208] Embodiment 4

[0209] Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of another lane line detection device based on artificial intelligence disclosed in an embodiment of the present invention. Among them, Figure 4 the described device can be applied to a management system, such as a local server or a cloud server for management, etc., and the embodiments of the present invention do not make limitations. As Figure 4 shown, the device may include:

[0210] A memory 301 storing executable program code;

[0211] A processor 302 coupled to the memory 301;

[0212] The processor 302 calls the executable program code stored in the memory 301 to execute the steps in the lane line detection method based on artificial intelligence described in Embodiment 1.

[0213] Embodiment 5

[0214] An embodiment of the present invention discloses a computer-readable storage medium that stores a computer program for electronic data exchange. Among them, the computer program causes a computer to execute the steps in the lane line detection method based on artificial intelligence described in Embodiment 1.

[0215] Embodiment 6

[0216] An embodiment of the present invention discloses a computer program product. The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps in the lane line detection method based on artificial intelligence described in Embodiment 1.

[0217] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place, or may be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0218] Through the specific descriptions of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, and the storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other medium that can be used to carry or store data and is computer-readable.

[0219] Finally, it should be noted that: The lane line detection method and device based on artificial intelligence disclosed in the embodiments of the present invention only disclose the preferred embodiments of the present invention, and are only used to illustrate the technical solutions of the present invention, rather than limiting them; Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence-based lane line detection method, characterized in that, The method includes: Obtaining the road image information to be detected; Preprocessing the road image information to be detected to obtain the target processing path information; Performing detection processing on the target processing path information to obtain the target lane line detection result information.

2. The lane line detection method based on artificial intelligence according to claim 1, wherein The performing detection processing on the target processing path information to obtain the target lane line detection result information includes: Performing feature extraction processing on the target processing path information by using a target feature extraction model to obtain target road feature information; Performing detection processing on the target road feature information by using a target detection model to obtain the target lane line detection result information.

3. The lane line detection method based on artificial intelligence according to claim 2, wherein The target feature extraction model includes a first convolution module, a first pooling module, a first extraction module, a second extraction module, a third extraction module, a fourth extraction module, and a fifth extraction module; wherein, The input end of the first convolution module is configured to receive the first model input of the target feature extraction model, and the output end of the first convolution module is connected to the input end of the first pooling module; the output end of the first pooling module is connected to the input end of the first extraction module; the output end of the first extraction module is respectively connected to the input ends of the second extraction module and the fifth extraction module; the output end of the second extraction module is respectively connected to the input ends of the third extraction module and the fifth extraction module; the output end of the third extraction module is respectively connected to the input ends of the fourth extraction module and the fifth extraction module; the output end of the fourth extraction module is connected to the input end of the fifth extraction module; the output end of the fifth extraction module is configured to output the first model output and the second model output of the target feature extraction model.

4. The lane line detection method based on artificial intelligence according to claim 3, characterized in that, The first extraction module includes a first convolution unit 4031, a second convolution unit 4032, a third convolution unit 4035, a fourth convolution unit 4038, a fifth convolution unit 40311, a sixth convolution unit 40314, a seventh convolution unit 40317, an eighth convolution unit 40320, a ninth convolution unit 40323, a first activation unit 4034, a second activation unit 4037, a third activation unit 40303, a fourth activation unit 40316, a fifth activation unit 40319, a sixth activation unit 40325, a first normalization unit 4033, a second normalization unit 4036, a third normalization unit 4039, a fourth normalization unit 40312, a fifth normalization unit 40315, a sixth normalization unit 40318, a seventh normalization unit 40321, an eighth normalization unit 40324, a first fusion unit 40310, and a second fusion unit 40322; wherein, The input end of the first convolution unit is configured as the input end of the first extraction module, and the output end of the first convolution unit is respectively connected to the input end of the second convolution unit, the input end of the fifth convolution unit, and the input end of the first fusion unit; the output end of the second convolution unit is connected to the input end of the first normalization unit; the output end of the first normalization unit is connected to the input end of the first activation unit; the output end of the first activation unit is connected to the input end of the third convolution unit; the output end of the third convolution unit is connected to the input end of the second normalization unit; the output end of the second normalization unit is connected to the input end of the second activation unit; the output end of the second activation unit is connected to the input end of the fourth convolution unit; the output end of the fourth convolution unit is connected to the input end of the third normalization unit; the output end of the third normalization unit is connected to the input end of the first fusion unit; the output end of the fifth convolution unit is connected to the input end of the fourth normalization unit; the output end of the fourth normalization unit is connected to the input end of the first fusion unit; the output end of the first fusion unit is connected to the input end of the third activation unit; the output end of the third activation unit is connected to the input end of the sixth convolution unit, the input end of the ninth convolution unit, and the input end of the second fusion unit; the output end of the sixth convolution unit is connected to the input end of the fifth normalization unit; the output end of the fifth normalization unit is connected to the input end of the fourth activation unit; the output end of the fourth activation unit is connected to the input end of the seventh convolution unit; the output end of the seventh convolution unit is connected to the input end of the sixth normalization unit; the output end of the sixth normalization unit is connected to the input end of the fifth activation unit; the output end of the fifth activation unit is connected to the input end of the eighth convolution unit; the output end of the eighth convolution unit is connected to the input end of the seventh normalization unit; the output end of the seventh normalization unit is connected to the output end of the second fusion unit; the output end of the ninth convolution unit is connected to the input end of the eighth normalization unit; the output end of the eighth normalization unit is connected to the input end of the second fusion unit; the output end of the second fusion unit is connected to the input end of the sixth activation unit; the output end of the sixth activation unit is configured as the output end of the first extraction module.

5. The lane line detection method based on artificial intelligence according to claim 3, characterized in that, The fifth extraction module includes a tenth convolution unit, an eleventh convolution unit, a twelfth convolution unit, a thirteenth convolution unit, a fourteenth convolution unit, a fifteenth convolution unit, a sixteenth convolution unit, a seventeenth convolution unit, an eighteenth convolution unit, a third fusion unit, a fourth fusion unit, a fifth fusion unit, a sixth fusion unit, a first sampling unit, a second sampling unit, a third sampling unit, a seventh activation unit, an eighth activation unit, a ninth normalization unit, a tenth normalization unit, a first slicing unit, a dimension transformation unit, and a multi-head attention unit; wherein, The input end of the tenth convolutional unit is connected to the output end of the first extraction module, and the output end of the tenth convolutional unit is connected to the input end of the fifth fusion unit; the input end of the eleventh convolutional unit is connected to the output end of the second extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the thirteenth convolutional unit; the output end of the thirteenth convolutional unit is connected to the input end of the third fusion unit; the input end of the twelfth convolutional unit is connected to the output end of the third extraction module, and the output end of the eleventh convolutional unit is connected to the input end of the fourteenth convolutional unit; the output end of the fourteenth convolutional unit is connected to the input end of the fourth fusion unit; the input end of the seventeenth convolutional unit is connected to the output end of the fourth extraction module, and the output end of the seventeenth convolutional unit is connected to the input end of the ninth normalization unit; the output end of the ninth normalization unit is connected to the input end of the seventh activation unit; the output end of the seventh activation unit is respectively connected to the input end of the multi-head attention unit and the input end of the first slicing unit; the output end of the multi-head attention unit is connected to the input end of the sixth fusion unit; the output end of the first slicing unit is connected to the input end of the dimension conversion unit; the output end of the dimension conversion unit is connected to the input end of the sixth fusion unit; the output end of the sixth fusion unit is connected to the input end of the eighteenth convolutional unit; the output end of the eighteenth convolutional unit is connected to the input end of the tenth normalization unit; the output end of the tenth normalization unit is connected to the input end of the eighth activation unit; the output end of the eighth activation unit is connected to the input end of the third sampling unit; the output end of the third sampling unit is connected to the input end of the fourth fusion unit; the output end of the fourth fusion unit is respectively connected to the input end of the second sampling unit and the input end of the fifteenth convolutional unit; the output end of the second sampling is connected to the input end of the third fusion unit; the output end of the third fusion unit is connected to the input end of the first sampling unit; the output end of the first sampling unit is connected to the input end of the fifth fusion unit; the output end of the fifth fusion unit is connected to the input end of the sixteenth convolutional unit; the output end of the sixteenth convolutional unit is configured to output the first model output of the target feature extraction model; the output end of the fifteenth convolutional unit is configured to output the second model output of the target feature extraction model.

6. The lane line detection method based on artificial intelligence according to claim 2, characterized in that, The target detection model includes a second convolutional module, a third convolutional module, a fourth convolutional module, a fifth convolutional module, a sixth convolutional module, a seventh convolutional module, an eighth convolutional module, a ninth convolutional module, a tenth convolutional module, a first activation module, a second activation module, a third activation module, a fourth activation module, a second pooling module, a third pooling module, a fourth pooling module, a fifth pooling module, a first network module and a second network module; wherein, The input end of the second convolution module and the input end of the third convolution module are configured to receive the output of the first model; the output end of the second convolution module is connected to the input end of the second pooling module; the output end of the second pooling module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the fourth convolution module; the output end of the fourth convolution module is connected to the input end of the first network module; the output end of the first network module is respectively connected to the input end of the second network module and the input end of the sixth convolution module; the output end of the third convolution module is connected to the input end of the third pooling module; the output end of the third pooling module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fifth convolution module; the output end of the fifth convolution module is connected to the input end of the second network module; the output end of the second network module is connected to the input end of the seventh convolution module; the input end of the eighth convolution module is configured to receive the output of the second model; the output end of the eighth convolution module is connected to the input end of the fourth pooling module; the output end of the fourth pooling module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the ninth convolution module; the output end of the ninth convolution module is connected to the input end of the fifth pooling module; the output end of the fifth pooling module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the tenth convolution module; the output end of the tenth convolution module is respectively connected to the input end of the sixth convolution module and the input end of the seventh convolution module; the output end of the sixth convolution module is configured to output the third model output of the target detection model; the output end of the seventh convolution module is configured to output the fourth model output of the target detection model.

7. The lane line detection method based on artificial intelligence according to claim 1, wherein The preprocessing of the road image information to be detected to obtain the target processing path information includes: Performing noise reduction and enhancement processing on the road image information to be detected to obtain the first processed road information; Performing color space conversion processing on the first processed road information to obtain the second processed road information; Performing image cropping processing on the second processed road information to obtain the target processed road information.

8. An artificial intelligence-based lane line detection device, characterized in that, The device includes: An acquisition module for acquiring road image information to be detected; A first processing module for preprocessing the road image information to be detected to obtain target processing path information; A second processing module for performing detection processing on the target processing path information to obtain target lane line detection result information.

9. A lane line detection device based on artificial intelligence, characterized in that, The device includes: A memory storing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory and executes the artificial intelligence-based lane line detection method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when called, are used to execute the artificial intelligence-based lane line detection method according to any one of claims 1-7.

Images