Self-adaptive lane line detection method, computer equipment and storage medium
Through the lane line detection method of adaptively allocating anchor points, the problem of low detection accuracy of curved lane line is solved, and accurate detection of multiple lane line types is achieved, improving the accuracy of detection.
Patent Information
- Application Number
- CN202411813498.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-10
- Publication Date
- 2025-04-25
AI Technical Summary
The existing lane line detection algorithm has low accuracy when detecting curved lane lines, making it difficult to effectively detect multiple lane line types.
Adaptive lane line detection method is adopted to extract lane line semantic information through feature extraction network, and row anchor points and column anchor points are adaptively allocated according to the detected information to perform lane line detection.
The detection accuracy of curved lane lines is improved, accurate judgment of multiple lane lines is achieved, and the accuracy of lane line detection is improved.
Smart Images
Figure CN119888653A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lane line detection, and in particular to an adaptive lane line detection method, a computer device and a storage medium. Background Art
[0002] Lane detection is an important function of autonomous vehicles in environmental perception, providing vehicles with accurate lane line shape and position information. Lane lines are small, narrow, and occupy a very low proportion in the image; road conditions and environment are complex and changeable, and the lane line shape has an inherent slenderness ratio; lane lines are interfered by other objects and lighting, such as guardrails, railways, telephone poles, sidewalks, buildings, etc.; at the same time, lane lines have a wide variety of line types, and often have certain occlusions and stains, which require imagination for occlusion. These factors result in low accuracy of lane detection algorithms.
[0003] Existing lane detection methods based on deep learning include lane detection algorithms based on semantic segmentation, such as LaneNet and Scnn. The model size is large and it is difficult to achieve real-time detection. Lane detection algorithms based on anchor detection, such as Ufld and LaneATT, use uniform anchor allocation, and fewer anchor points are allocated to lanes with large curvature, resulting in poor detection accuracy of curved lanes. The current detection algorithm is difficult to effectively detect a variety of lane line types. Summary of the invention
[0004] In view of this, the present invention provides an adaptive lane line detection method, a computer device and a storage medium to solve the problem of low detection accuracy of curved lane lines.
[0005] In a first aspect, the present invention provides an adaptive lane line detection method, which comprises: acquiring a target lane image including multi-line lane lines; in a feature extraction module, using a feature extraction network to perform feature extraction on the target lane image to obtain lane line semantic information of the multi-line lane lines; in an anchor point allocation module, using the lane line semantic information to adaptively allocate row anchor points and column anchor points to the target lane image; in an existence branch module, performing lane line existence detection on the row anchor points and the column anchor points to obtain the row anchor points and the column anchor points of the existing lane lines; in a detection head module, using the row anchor points and the column anchor points of the existing lane lines to obtain lane line data.
[0006] In this implementation, lane line features are extracted from the target lane image, and anchor points are further customized according to the detected lane line information, and lane line detection is performed based on the assigned anchor points. Among them, this adaptive anchor point allocation method can pay more attention to the semantic information of the lane line, improve the detection accuracy of the curved lane line, and achieve the problem of accurate identification of various lane line types.
[0007] In one implementation, a feature extraction network is used to extract features from a target lane image to obtain lane line semantic information of a multi-line lane line, including: using a multi-layer perceptron to extract features from the lane line semantic information to obtain global semantic information of the multi-line lane line; using the global semantic information to generate a row anchor point allocation function and a column anchor point allocation function; and assigning row anchor points and column anchor points to the target lane image based on the row anchor point allocation function and the column anchor point allocation function.
[0008] In this implementation, a row anchor point allocation function and a column anchor point allocation function are constructed to allocate row anchor points and column anchor points respectively, which can adaptively allocate anchor points according to different lane lines, improve the accuracy of anchor point allocation, and further improve the accuracy of lane line detection.
[0009] In one implementation, allocating row anchor points and column anchor points to the target lane image based on the row anchor point allocation function and the column anchor point allocation function includes: Among them, F row is the number of row anchor points, Assign function to row anchor points, F col is the number of column anchor points, is the column anchor point allocation function, Δy is the interval coordinate of the row anchor points along the height direction, Δx is the interval coordinate of the column anchor points along the width direction, m and n are the number of calculations allocated to the lane line and the train line, respectively, k1 and k2 are the highest order of adaptive learning; w ij is the weight of the jth order of the i-th lane line.
[0010] In one implementation, the closer Δy is to the height top of the target lane image, the more row anchor points are allocated by the row anchor point allocation function; and the closer Δx is to the width center of the target lane image, the more column anchor points are allocated by the column anchor point allocation function.
[0011] In this implementation, row anchor points are allocated according to the upper and lower positions of the target lane image, and column anchor points are allocated according to the left and right positions of the target lane image, which can better identify curve information and improve the accuracy of curve detection.
[0012] In one implementation, the lane line data includes the position data of the lane line, and obtaining the lane line data using the row anchor points and column anchor points of the existing lane lines includes: in the positioning branch module, using the positioning detection algorithm to locate and detect the row anchor points and the column anchor points to obtain the position information of each row anchor point and the position information of each column anchor point; in the detection head module, based on the position information, determining the position information of the row anchor points and the column anchor points of the existing lane lines to obtain the position data of the lane lines.
[0013] In this implementation, the positioning branch module is used to further identify the lane line position information to improve the lane line detection accuracy.
[0014] In one implementation, the adaptive lane line detection method further includes: in a line type discrimination module, extracting features of the lane line semantic information to obtain line type features of the lane line; and obtaining a corresponding lane line type based on the line type features.
[0015] In this implementation, a line type discrimination module is used to further identify line type features, thereby improving the comprehensiveness of subsequent line type detection.
[0016] In one implementation, the lane line data includes lane line type data, and obtaining the lane line data based on the row anchor points and column anchor points of the existing lane lines includes: in the detection head module, based on the lane line type, determining the type information of the row anchor points and column anchor points of the existing lane lines, and obtaining the lane line type data.
[0017] In one implementation, the adaptive lane line detection method further includes: in the detection head module, combining the row anchor points and column anchor points of the lane lines, the type data of the lane lines, and the position data of the lane lines, and outputting the lane line detection results.
[0018] In a second aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the adaptive lane line detection method of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.
[0019] In a third aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the adaptive lane line detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 is a schematic structural diagram of an adaptive lane detection device according to an embodiment of the present invention;
[0022] Figure 2 is a flow chart of an adaptive lane detection method according to an embodiment of the present invention;
[0023] Figure 3 is a flow chart of another adaptive lane detection method according to an embodiment of the present invention;
[0024] Figure 4 is a schematic structural diagram of another adaptive lane detection device according to an embodiment of the present invention;
[0025] Figure 5 is a flowchart of another adaptive lane detection method according to an embodiment of the present invention;
[0026] Figure 6 is a schematic structural diagram of another adaptive lane detection device according to an embodiment of the present invention;
[0027] Figure 7 is a flowchart of another adaptive lane detection method according to an embodiment of the present invention;
[0028] Figure 8 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0030] In this embodiment, an adaptive lane line detection device is provided. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.
[0031] This embodiment provides an adaptive lane detection device. Figure 1 , Figure 1 Schematic diagram of the structure of an adaptive lane detection device according to an embodiment of the present invention.
[0032] The adaptive lane line detection device comprises a feature extraction module, an anchor point allocation module, an existence branch module and a detection head module. The feature extraction module is connected to the anchor point allocation module, the anchor point allocation module is connected to the existence branch module, and the existence branch module is connected to the detection head module.
[0033] Among them, the feature extraction module is used to extract features from the target lane image to obtain the lane semantic information of the multi-line lane. The anchor point allocation module is used to adaptively allocate row anchor points and column anchor points to the target lane image using the lane semantic information. The existence branch module is used to perform lane existence detection on the row anchor points and column anchor points to obtain the row anchor points and column anchor points of the existing lane lines. The detection head module is used to obtain the lane data using the row anchor points and column anchor points of the existing lane lines.
[0034] According to an embodiment of the present invention, an embodiment of an adaptive lane line detection method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0035] In this embodiment, an adaptive lane line detection method is provided. Figure 2 is a flow chart of an adaptive lane detection method according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not based on Figure 2 The process sequence shown is limited. Figure 2 As shown, the process includes the following steps:
[0036] Step S201, acquiring a target lane image including multi-line lane lines.
[0037] An image sensor is used to collect a target lane image of the current road, wherein the target lane image includes multi-line lane lines.
[0038] Among them, multi-line lane lines include lane lines of multiple different colors and line types, such as white and yellow, dotted lines and solid lines, etc.
[0039] In one implementation, an image collector on the vehicle is used to collect an image of the target lane ahead in real time.
[0040] Step S202: In the feature extraction module, a feature extraction network is used to extract features of the target lane image to obtain lane line semantic information of the multi-line lane line.
[0041] The target lane image is input into the feature extraction module, and the feature extraction module is used to extract the initial features of the lane line to facilitate the subsequent further identification of the lane line information.
[0042] In one implementation, a residual network is used to extract lane line semantic information of multi-line lane lines.
[0043] Step S203: In the anchor point allocation module, the lane line semantic information is used to adaptively allocate row anchor points and column anchor points to the target lane image.
[0044] The lane line semantic information is input into the anchor point assignment module, and the row anchor points and column anchor points are adaptively assigned in the target lane image according to the point set assignment features of the lane line.
[0045] Specifically, anchor points are assigned to lane lines according to their positions in the target lane image, such as lane lines at the top of the target lane image are assigned more row anchor points, and lane lines at the center of the target lane image are assigned more column anchor points.
[0046] Step S204: In the existence branch module, lane line existence detection is performed on the row anchor points and the column anchor points to obtain the row anchor points and the column anchor points where the lane lines exist.
[0047] The allocated row anchor points and column anchor points are input into the branch module, and each anchor point is detected to identify whether there is a lane line in the anchor point.
[0048] Specifically, the existence branch module outputs the probability of each lane line existing at each assigned anchor point.
[0049] Step S205: In the detection head module, lane line data is obtained using the row anchor points and column anchor points of the existing lane lines.
[0050] The row anchor points, column anchor points and corresponding data on whether there are lane lines are input into the detection head module, and finally the lane line detection result data is output.
[0051] The adaptive lane line detection method provided in this embodiment extracts lane line features from the target lane image, further customizes anchor points according to the detected lane line information, and performs lane line detection based on the allocated anchor points. The adaptive anchor point allocation method can pay more attention to the semantic information of the lane line, improve the detection accuracy of the curved lane line, and realize the problem of accurate identification of various lane line types.
[0052] In this embodiment, another adaptive lane line detection method is provided. Figure 3 is a flowchart of another adaptive lane detection method according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not based on Figure 3 The process sequence shown is limited. Figure 3 As shown, the process includes the following steps:
[0053] Step S301, acquiring a target lane image including multi-line lane lines.
[0054] Step S302: In the feature extraction module, a feature extraction network is used to extract features of the target lane image to obtain lane line semantic information of the multi-line lane line.
[0055] Specifically, a multi-layer perceptron is used to extract features of lane line semantic information to obtain the global semantic information of multi-line lane lines.
[0056] In one implementation, the feature extraction module includes a lightweight residual network of resnet18. The residual network is used to extract high-level semantic features of lane lines in the target lane image. Specifically, the residual network includes four residual modules connected in series, the input is the target lane image, and the output lane line semantic information is a high-level semantic feature map F1, which is used as input data of the anchor point allocation module.
[0057] Step S303: In the anchor point allocation module, the lane line semantic information is used to adaptively allocate row anchor points and column anchor points to the target lane image.
[0058] Specifically, the above step S303 includes:
[0059] Step S3031: Generate a row anchor point allocation function and a column anchor point allocation function using global semantic information.
[0060] Specifically, the construction principle of the row anchor point allocation function is: the closer Δy is to the height top of the target lane image, the more row anchor points the row anchor point allocation function allocates; the construction principle of the column anchor point allocation function is: the closer Δx is to the width center of the target lane image, the more column anchor points the column anchor point allocation function allocates.
[0061] Step S3032: assigning row anchor points and column anchor points to the target lane image based on the row anchor point assignment function and the column anchor point assignment function.
[0062] Specifically, row anchor points and column anchor points are assigned to the target lane image based on the row anchor point assignment function and the column anchor point assignment function:
[0063]
[0064] Among them, F row is the number of row anchor points, Assign function to row anchor points, F col is the number of column anchor points, f col k2 (Δx) is the column anchor point allocation function, Δy is the interval coordinate of the row anchor points along the height direction, Δx is the interval coordinate of the column anchor points along the width direction, m and n are the number of calculations allocated to the lane line and the train line, respectively, k1 and k2 are the highest order of adaptive learning; w ij is the weight of the jth order of the i-th lane line.
[0065] Step S304: In the existence branch module, lane line existence detection is performed on the row anchor points and the column anchor points to obtain the row anchor points and the column anchor points where the lane lines exist.
[0066] In the positioning branch module, the row anchor points and column anchor points are positioned and detected using a positioning detection algorithm to obtain the position information of each row anchor point and the position information of each column anchor point.
[0067] Specifically, the input data of the existence branch module is the information of each row anchor point and each column anchor point. Through convolution and dimensionality conversion, the probability of the existence of each assigned anchor point for each lane line is output to obtain the row anchor point and column anchor point of the existing lane line.
[0068] Step S305: In the detection head module, lane line data is obtained using the row anchor points and column anchor points where lane lines exist.
[0069] In the detection head module, based on the position information, the position information of the row anchor points and column anchor points where the lane lines exist is determined to obtain the position data of the lane lines.
[0070] In this implementation, row anchor point allocation function and column anchor point allocation function are constructed to allocate row anchor points and column anchor points respectively, which can adaptively allocate anchor points according to different lane lines, improve the accuracy of anchor point allocation, and further improve the accuracy of lane line detection. Specifically, row anchor points are allocated according to the upper and lower positions of the target lane image, and column anchor points are allocated according to the left and right positions of the target lane image, which can better identify curve information and improve the accuracy of curve detection.
[0071] This embodiment provides another adaptive lane detection device. Figure 4 , Figure 4 is a schematic structural diagram of another adaptive lane detection device according to an embodiment of the present invention.
[0072] The adaptive lane line detection device includes a feature extraction module, an anchor point allocation module, an existence branch module, a positioning branch module and a detection head module.
[0073] The feature extraction module is connected to the anchor point allocation module, the anchor point allocation module is connected to the existence branch module and the positioning branch module, and the existence branch module and the positioning branch module are connected to the detection head module.
[0074] Among them, the positioning branch module is used to use the positioning detection algorithm to perform positioning detection on the row anchor points and the column anchor points to obtain the position information of each row anchor point and the position information of each column anchor point.
[0075] In this embodiment, another adaptive lane line detection method is provided. Figure 5 is a flow chart of an adaptive lane detection method according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not based on Figure 5 The process sequence shown is limited. Figure 5 As shown, the process includes the following steps:
[0076] Step S501, acquiring a target lane image including multi-line lane lines.
[0077] Step S502: In the feature extraction module, a feature extraction network is used to extract features from the target lane image to obtain lane line semantic information of the multi-line lane line.
[0078] Step S503: In the anchor point allocation module, the lane line semantic information is used to adaptively allocate row anchor points and column anchor points to the target lane image.
[0079] Step S504: In the existence branch module, lane line existence detection is performed on the row anchor points and the column anchor points to obtain the row anchor points and the column anchor points where the lane lines exist.
[0080] Step S505: In the positioning branch module, a positioning detection algorithm is used to perform positioning detection on the row anchor points and the column anchor points to obtain the position information of each row anchor point and the position information of each column anchor point.
[0081] The data of the allocated row anchor points and column anchor points are input into the positioning branch module, and the position information of each anchor point is determined using the positioning detection algorithm.
[0082] Specifically, the anchor point position coordinate probabilities calculated by the row and column anchor points are output respectively, and through convolution and dimensionality conversion, their probabilities are output as the position coordinates for determining the existence of the lane line anchor point.
[0083] Step S506: In the detection head module, the lane line data is obtained using the row anchor points and column anchor points of the existing lane lines.
[0084] The row anchor points, column anchor points, data corresponding to whether there are lane lines, and the position coordinate data of each anchor point are input into the detection head module. Based on the position information, the position information of the row anchor points and column anchor points where the lane lines exist is determined to obtain the position data of the lane lines.
[0085] The adaptive lane line detection method provided in this embodiment uses a positioning branch module to further identify lane line position information, thereby improving the accuracy of lane line detection.
[0086] This embodiment provides another adaptive lane detection device. Figure 6 , Figure 6 4 is a schematic diagram of the structure of another adaptive lane detection device according to an embodiment of the present invention.
[0087] The adaptive lane line detection device includes an image processing module, a feature extraction module, an anchor point allocation module, an existence branch module, a positioning branch module, a line type discrimination module and a detection head module.
[0088] The feature extraction module is connected to the anchor point allocation module and the line type discrimination module, the anchor point allocation module is connected to the existence branch module and the positioning branch module, and the existence branch module, the line type discrimination module and the positioning branch module are connected to the detection head module.
[0089] Among them, the line type discrimination module is used to extract features of the lane line semantic information, obtain the line type features of the lane line, and obtain the corresponding lane line type based on the line type features.
[0090] In this embodiment, another adaptive lane line detection method is provided. Figure 7 is a flow chart of an adaptive lane detection method according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not based on Figure 7 The process sequence shown is limited. Figure 7 As shown, the process includes the following steps:
[0091] Step S701, acquiring a target lane image including multi-line lane lines.
[0092] Step S702: In the image processing module, pre-process the target lane image.
[0093] Among them, the image processing module includes image ROI area cropping, image scaling, image enhancement and other processing.
[0094] Specifically, the target lane image is input into the image processing module, the target lane image is cropped, the ROI area containing the lane line is cropped, and the interfering background such as the sky above the lane line is removed. At the same time, the target lane image is scaled to a fixed resolution to facilitate subsequent unified input into the model for calculation.
[0095] The image processing module is trained before actual use. During the model training process, the data needs to be enhanced. Data enhancement includes data radiation transformation, contrast adjustment, and noise addition to increase data diversity and improve the robustness of the model.
[0096] Step S703: In the feature extraction module, a feature extraction network is used to extract features of the target lane image to obtain lane line semantic information of the multi-line lane line.
[0097] The fixed-resolution target lane image is input into the backbone feature extraction module. The backbone feature extraction module adopts the resnet18 lightweight residual network and connects 4 residual blocks in series. Each residual block consists of a feature extraction structure and a residual structure. The final output high-level semantic feature map contains high-level semantic information of the lane line.
[0098] Step S704: In the anchor point allocation module, the lane line semantic information is used to adaptively allocate row anchor points and column anchor points to the target lane image.
[0099] Before actual use, the model of the anchor point assignment module is trained. According to the input data set, the lane line ID calculated by specifying the row and column anchor points is adaptively learned to learn the weights of the rows and columns, and the relationship between the number of anchor points and the intervals corresponding to the rows and columns is learned.
[0100] Specifically, the high-level semantic information extracted from the backbone feature extraction module is input into the anchor point allocation module. First, the features are flattened along the height and width directions, and the flattened features are input into the multi-layer perceptron to extract the global semantic information of the lane line. For each input feature map, the number of anchor points is adaptively allocated according to rows and columns. The specific allocation method refers to step S303 and will not be described here.
[0101] Step S705: In the existence branch module, lane line existence detection is performed on the row anchor points and the column anchor points to obtain the row anchor points and the column anchor points where the lane lines exist.
[0102] Step S706: In the positioning branch module, a positioning detection algorithm is used to perform positioning detection on the row anchor points and the column anchor points to obtain the position information of each row anchor point and the position information of each column anchor point.
[0103] The assigned anchor points are input into the existence branch and the positioning branch respectively. Through convolutional feature extraction and dimensionality conversion, the existence probability and positioning probability corresponding to each anchor point are learned.
[0104] Step S707, in the line type discrimination module, feature extraction is performed on the lane line semantic information to obtain the line type features of the lane line, and the corresponding lane line type is obtained based on the line type features.
[0105] The high-level semantic information extracted from the backbone feature extraction module is input into the line type discrimination module, and the lane line type features in the high-level semantic information are further extracted through convolution, layer normalization, activation and full connection.
[0106] Among them, the lane line type features include high-level semantic features such as different colors and line types.
[0107] Step S708: In the detection head module, the lane line data is obtained using the row anchor points and column anchor points of the existing lane lines.
[0108] In the detection head module, the lane line detection result is output by combining the row anchor points and column anchor points of the lane line, the type data of the lane line, and the position data of the lane line.
[0109] Specifically, the intersection of the output of the line type discrimination module and the output of the existence branch module, and the intersection of the output of the positioning branch module and the output of the existence branch module are respectively calculated, and the final lane line detection result with coordinate information and line type category is output.
[0110] The adaptive lane detection method provided in this embodiment uses a line type recognition module to further identify line type features, thereby improving the comprehensiveness of subsequent line type detection.
[0111] The adaptive lane line detection device provided in this embodiment is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here for what has been described.
[0112] The adaptive lane line detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0113] The further functional description of each of the above modules is the same as that of the above corresponding embodiments and will not be repeated here.
[0114] The embodiment of the present invention also provides a computer device having the above Figure 1 Or the adaptive lane line detection device shown in 4 or 6.
[0115] See also Figure 8 , Figure 8 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 8 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.
[0116] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0117] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.
[0118] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0119] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0120] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 8 The example of connecting through bus is taken in the following.
[0121] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.
[0122] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0123] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.
[0124] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. An adaptive lane detection method, characterized in that: The method comprises: Acquire a target lane image including multi-line lane lines; In the feature extraction module, a feature extraction network is used to extract features of the target lane image to obtain lane line semantic information of the multi-line lane line; In an anchor point allocation module, row anchor points and column anchor points are adaptively allocated to the target lane image using the lane line semantic information; In the existence branch module, lane line existence detection is performed on the row anchor point and the column anchor point to obtain the row anchor point and the column anchor point where the lane line exists; In the detection head module, lane line data is obtained using the row anchor points and the column anchor points where lane lines exist.
2. The adaptive lane detection method according to claim 1, characterized in that: The extracting features of the target lane image using a feature extraction network to obtain lane line semantic information of the multi-line lane line includes: Using a multi-layer perceptron to extract features from the lane line semantic information, to obtain global semantic information of the multi-line lane line; Generate a row anchor point allocation function and a column anchor point allocation function using the global semantic information; The row anchor point and the column anchor point are allocated to the target lane image based on the row anchor point allocation function and the column anchor point allocation function.
3. The adaptive lane detection method according to claim 2, characterized in that: The allocating the row anchor point and the column anchor point to the target lane image based on the row anchor point allocation function and the column anchor point allocation function comprises: Among them, F row is the number of row anchor points, Assign a function to the row anchor point, F col is the number of column anchor points, is the column anchor point allocation function, Δy is the interval coordinate of the row anchor points along the height direction, Δx is the interval coordinate of the column anchor points along the width direction, m and n are the number of calculations allocated to the lane line and the train line, respectively, k1 and k2 are the highest order of adaptive learning; w ij is the weight of the jth order of the i-th lane line.
4. The adaptive lane detection method according to claim 3, characterized in that: The closer the Δy is to the height top of the target lane image, the more row anchor points are allocated by the row anchor point allocation function; the closer the Δx is to the width center of the target lane image, the more column anchor points are allocated by the column anchor point allocation function.
5. The adaptive lane detection method according to claim 1, characterized in that: The lane line data includes the position data of the lane line, and obtaining the lane line data by using the row anchor point and the column anchor point where the lane line exists includes: In the positioning branch module, the row anchor points and the column anchor points are positioned and detected by using a positioning detection algorithm to obtain position information of each row anchor point and position information of each column anchor point; In the detection head module, based on the position information, the position information of the row anchor point and the column anchor point where the lane line exists is determined to obtain the position data of the lane line.
6. The adaptive lane detection method according to claim 5, characterized in that: The method further comprises: In the line type discrimination module, feature extraction is performed on the lane line semantic information to obtain the line type feature of the lane line; The corresponding lane line type is obtained based on the line type features.
7. The adaptive lane detection method according to claim 6, characterized in that: The lane line data includes lane line type data, and the lane line data obtained based on the row anchor point and the column anchor point of the existing lane line includes: In the detection head module, based on the lane line type, type information of the row anchor point and the column anchor point where the lane line exists is determined to obtain type data of the lane line.
8. The adaptive lane detection method according to claim 7, characterized in that: The method further comprises: In the detection head module, the lane line detection result is output by combining the row anchor point and the column anchor point where the lane line exists, the type data of the lane line, and the position data of the lane line.
9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the adaptive lane line detection method according to any one of claims 1 to 8 by executing the computer instructions.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the adaptive lane line detection method according to any one of claims 1 to 8.