CN112686080A - Method and device for detecting lane line - Google Patents
Method and device for detecting lane line Download PDFInfo
- Publication number
- CN112686080A CN112686080A CN201910990018.3A CN201910990018A CN112686080A CN 112686080 A CN112686080 A CN 112686080A CN 201910990018 A CN201910990018 A CN 201910990018A CN 112686080 A CN112686080 A CN 112686080A
- Authority
- CN
- China
- Prior art keywords
- lane line
- layer
- network
- resnet
- version
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000001514 detection method Methods 0.000 claims abstract description 62
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 238000004590 computer program Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 abstract description 17
- 238000005286 illumination Methods 0.000 abstract description 13
- 238000010586 diagram Methods 0.000 description 31
- 238000013528 artificial neural network Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 239000003086 colorant Substances 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 241000282979 Alces alces Species 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000003550 marker Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000005452 bending Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Abstract
The invention discloses a method and a device for detecting lane lines, and relates to the technical field of computers. One embodiment of the method comprises: obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; feature recognition is performed on the extracted features using a full convolutional network for lane line detection. The implementation method can solve the problem of low detection result accuracy caused by too simple feature extraction bottom layer model; the method can ensure the processing performance of the network model and improve the accuracy of lane line detection while using the complex network model to perform feature extraction and feature identification, so that higher accuracy can be obtained under complex scenes with larger changes such as illumination and the like.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a method and a device for detecting lane lines.
Background
Lane line detection is of great significance to autonomous driving. The traditional lane line detection method has poor effect in complex road scenes or scenes with large illumination changes such as night and rainy days due to the characteristics of manual design. In recent years, lane line detection methods based on a deep neural network are increasing. Compared with the traditional artificially involved features, the features extracted by the deep neural network have better robustness to illumination and the like, but the detection accuracy is poor due to the fact that models of the underlying neural network such as VGG and the like are simple.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method and an apparatus for lane line detection, which can solve the problem of low accuracy of a detection result due to an excessively simple bottom layer model for feature extraction; the method can ensure the processing performance of the network model and improve the accuracy of lane line detection while using the complex network model to perform feature extraction and feature identification, so that higher accuracy can be obtained under complex scenes with larger changes such as illumination and the like.
To achieve the above object, according to an aspect of an embodiment of the present invention, there is provided a method of performing lane line detection.
A method of lane line detection, comprising: obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; feature recognition is performed on the extracted features using a full convolutional network for lane line detection.
Optionally, obtaining the binary bitmap and the example graph of the lane line according to the original picture of the lane line includes: obtaining coordinates of a series of points on the lane line according to the original picture of the lane line; performing cubic spline interpolation on the series of points to obtain a lane line trajectory line; setting pixel values of points located on a track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line; the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example graph of the lane line.
Optionally, the feature identifying the extracted features using a full convolutional network comprises: selecting a decoding layer for feature recognition, wherein the decoding layer is a designated layer of the residual error network; and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network so as to perform feature identification.
Optionally, the residual network is ResNet-50, and the selected decoding layer is: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
Optionally, the residual network is MobileNet, and the selected decoding layer is: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
According to still another aspect of the embodiments of the present invention, there is provided an apparatus for lane line detection.
An apparatus for lane line detection, comprising: the image acquisition module is used for obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line; the characteristic extraction module is used for extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; and the characteristic identification module is used for carrying out characteristic identification on the extracted characteristics by using a full convolution network so as to detect the lane line.
Optionally, the image acquisition module is further configured to: obtaining coordinates of a series of points on the lane line according to the original picture of the lane line; performing cubic spline interpolation on the series of points to obtain a lane line trajectory line; setting pixel values of points located on a track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line; the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example graph of the lane line.
Optionally, the feature identification module is further configured to: selecting a decoding layer for feature recognition, wherein the decoding layer is a designated layer of the residual error network; and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network so as to perform feature identification.
Optionally, the residual network is ResNet-50, and the selected decoding layer is: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
Optionally, the residual network is MobileNet, and the selected decoding layer is: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
According to still another aspect of the embodiments of the present invention, there is provided an electronic apparatus that performs lane line detection.
An electronic device that performs lane line detection, comprising: one or more processors; the storage device is used for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method for detecting the lane line provided by the embodiment of the invention.
According to yet another aspect of embodiments of the present invention, a computer-readable medium is provided.
A computer-readable medium, on which a computer program is stored, which, when executed by a processor, implements the method for lane line detection provided by an embodiment of the present invention.
One embodiment of the above invention has the following advantages or benefits: obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; the extracted features are subjected to feature recognition by using a full convolution network so as to detect the lane lines, so that the problem of low accuracy of a detection result caused by the fact that a feature extraction bottom layer model is too simple is solved; the method can ensure the processing performance of the network model and improve the accuracy of lane line detection while using the complex network model to perform feature extraction and feature identification, so that higher accuracy can be obtained under complex scenes with larger changes such as illumination and the like.
Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
FIG. 1 is a schematic diagram of the main steps of a method for lane line detection according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a neural network for lane line detection according to an embodiment of the present invention;
FIG. 3(a) is a diagram illustrating an original image of a lane line according to a first embodiment of the present invention;
FIG. 3(b) is a binary bit diagrammatic illustration of a lane marker of the first embodiment of the present invention;
FIG. 3(c) is a schematic diagram of an example of a lane line according to the first embodiment of the present invention;
FIG. 4 is a diagram illustrating a process for decoding a decoding layer according to an embodiment of the present invention;
FIG. 5(a) is a diagram of an original picture according to a second embodiment of the present invention;
FIG. 5(b) is a diagram illustrating the detection result of the second embodiment of the present invention;
FIG. 6 is a diagram illustrating a process of decoding a decoding layer according to another embodiment of the present invention;
FIG. 7(a) is a diagram of an original picture according to a third embodiment of the present invention;
FIG. 7(b) is a diagram showing the detection result of the third embodiment of the present invention;
FIG. 8 is a schematic diagram of the main blocks of an apparatus for lane line detection according to an embodiment of the present invention;
FIG. 9 is an exemplary system architecture diagram in which embodiments of the present invention may be employed;
fig. 10 is a schematic block diagram of a computer system suitable for use in implementing a terminal device or server according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, in which various details of embodiments of the invention are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
In order to solve the problems in the prior art, the invention provides a method and a device for detecting lane lines, which can improve the accuracy of the result of detecting the lane lines in a complex scene with large changes of illumination and the like.
Fig. 1 is a schematic diagram of main steps of a method for lane line detection according to an embodiment of the present invention. As shown in fig. 1, the method for detecting a lane line according to the embodiment of the present invention mainly includes the following steps S101 to S103.
Step S101: obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line;
step S102: extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network;
step S103: feature recognition is performed on the extracted features using a full convolutional network for lane line detection.
According to an embodiment of the present invention, in step S101, obtaining the binary bitmap and the example map of the lane line according to the original image of the lane line may specifically include:
obtaining coordinates of a series of points on the lane line according to the original picture of the lane line;
carrying out cubic spline interpolation on the series of points to obtain a lane line trajectory;
setting pixel values of points located on the track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line;
the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example map of the lane line.
Pixels are small blocks that make up an image, each of these small blocks having a distinct location and assigned color value, the color and location of the small block determining what the image appears, and the pixel value representing the average luminance information of a particular small block. When the binary bitmap and the example graph of the lane line are obtained according to the original picture of the lane line, each point on the original picture corresponds to one pixel, so that the point can be called as a pixel point, and the color value allocated to the pixel point is described through the pixel value. Because the color of a pixel is represented by three values of RGB (RGB is a color representing three channels of red, green, and blue), the pixel value of a pixel is represented by a vector matrix corresponding to the three colors. In an embodiment of the present invention, for a point located on the trajectory line of the lane line, the pixel value corresponding to the point is represented as [255, 255, 255], corresponding to the values of the colors of the three channels, respectively.
According to another embodiment of the present invention, in step S103, the performing feature identification on the extracted features by using a full convolution network specifically includes:
selecting a decoding layer for carrying out feature identification, wherein the decoding layer is a designated layer of a residual error network;
and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network to perform feature identification.
In one embodiment of the present invention, the selected residual network is, for example, ResNet-50, and the selected decoding layers are: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
In another embodiment of the present invention, the selected residual network is MobileNet, and the selected decoding layers are: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
In order to realize the method for detecting the lane line, the invention provides a neural network, and the original picture, the binary bitmap and the example graph corresponding to the lane line are input into the neural network to realize the detection of the lane line.
Fig. 2 is a schematic structural diagram of a neural network for lane line detection according to an embodiment of the present invention. As shown in fig. 2, the neural network for lane line detection according to the embodiment of the present invention mainly includes two parts, i.e.,
the coding model uses a residual error network to perform feature extraction on an original picture, a binary bitmap and an example graph of an input lane line;
the decoding model uses a full convolution network to perform feature recognition on a decoding layer selected from the coding model so as to realize lane line detection.
The residual error network is a deep convolution network and is mainly applied to the scenes of image classification, detection, positioning and the like. The residual network is easier to optimize and can improve accuracy by adding a comparable depth. The core of the residual network is to solve the side effect (degradation problem) caused by increasing the depth, so that the network performance can be improved by simply increasing the network depth. In the embodiment of the invention, the residual error network is used for extracting the characteristics of the image data related to the lane line, so that the problem of low accuracy of the detection result caused by too simple bottom layer model of characteristic extraction can be solved.
In the embodiment of the invention, the coding model part is divided into two branches, one branch performs feature extraction through binary segmentation, the other branch performs feature extraction through instance segmentation, and the two branches share a residual error network. Therefore, the original data set input into the coding model includes not only the original picture of the lane line but also the binary bitmap and the instance map processed from the original picture.
According to the technical scheme of one embodiment of the invention, the binary bitmap and the example graph are obtained by the following steps:
obtaining coordinates of a series of points on the lane line according to the original picture of the lane line;
carrying out cubic spline interpolation on the series of points to obtain a lane line trajectory;
setting pixel values of points located on the track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line;
the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example map of the lane line.
Referring to fig. 3(a) -3 (c), fig. 3(a) is a schematic diagram of an original image of a lane line according to a first embodiment of the present invention; FIG. 3(b) is a binary bit diagrammatic illustration of a lane marker of the first embodiment of the present invention; fig. 3(c) is a schematic diagram of an example of the lane line according to the first embodiment of the present invention. Fig. 3(a) -3 (c) show the correspondence between the original picture, the binary bitmap and the example graph of the lane line according to the first embodiment of the present invention. Where fig. 3(a) is an original picture of a lane line, which shows a picture of a lane line taken during the driving of a vehicle, coordinates of a series of points on the lane line can be obtained from the original picture. Then, the trajectory line of the lane line can be obtained by performing cubic spline interpolation on the series of points. Cubic Spline Interpolation (Spline Interpolation) is abbreviated as Spline Interpolation, and is a process of obtaining a curve function set mathematically by solving a three bending moment equation set through a smooth curve of a series of shape value points.
Thereafter, by setting the pixel value of a point located on the lane line trajectory line to [255, 255, 255], the lane line trajectory line is made to appear white; by setting the pixel values of the points located at the remaining positions to [0, 0, 0], the other positions of the non-lane-line trajectory line are displayed in black, and a binary bitmap of the lane line is obtained, as shown in fig. 3 (b).
An example map of the lane lines is obtained by setting the pixel values of the points located on the different lane line trajectory lines to different levels other than 0, respectively, and setting the pixel values of the points located at the remaining positions to [0, 0, 0], as shown in fig. 3 (c). For example: by setting the pixel values of the points located on the different lane line trajectory lines to 5 levels of 20, 70, 120, 170, 220, respectively, it is possible to distinguish between lanes representing different types, for example: left lane, left lane of the vehicle, right lane, fifth lane, etc. The left lane of the self-vehicle is a lane line next to the left side of the self-vehicle, the right lane of the self-vehicle is a lane line next to the right side of the self-vehicle, and the fifth lane is not limited to be on the left side or the right side of the driving direction of the self-vehicle.
By using a cubic spline interpolation method to obtain the binary bitmap and the example graph of the lane line, the data result of the obtained binary bitmap and the example graph can be more accurate. After the original picture, binary bitmap and example map of the lane lines are obtained, they can be input into the coding model for feature extraction. And then decoding the selected decoding layer through a decoding model to perform feature recognition on the extracted features, thereby realizing the detection of the lane line.
According to an embodiment of the present invention, the residual network selected in the coding model is, for example, ResNet-50, and the decoding model part adopts a full Convolutional neural network fcn (full Convolutional neural networks). After a lot of data training and result analysis, when the selected decoding (decoder) layers are: the best effect is obtained when the resnet _ V1_50/block4 layer is represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
Fig. 4 is a diagram illustrating a process of decoding a decoding layer according to an embodiment of the present invention. As shown in fig. 4, the specific process of decoding the above decoding layers to realize feature identification is as follows:
1. processing the resnet _ v1_50/block4 layers, wherein the feature size of the resnet _ v1_50/block4 layers is 8 × 16 × 2048, and the feature size is 8 × 16 × 512, wherein 512 convolution kernels of 7 × 7 are utilized to obtain a fully connected layer fc6(full convolution) layer; then, 512 convolution kernels of 1 × 1 are used for the fc6 layer to obtain an fc7 (fusion convolution) layer with the characteristic size of 8 × 16 × 512; then, 64 convolution kernels of 1 × 1 are applied to fc7 layers to obtain score _ orignal layers with the characteristic size of 8 × 16 × 64; finally, carrying out deconvolution (deconv) operation on the score _ original layer by using 256 convolution kernels with 4 × 4 to obtain a deconv _1 layer with the characteristic size of 16 × 32 × 64;
2. processing a resnet _ v1_50/block3/unit _ 1/botteleck _ v1 layer, wherein the characteristic size of the layer is 16 × 32 × 1024, firstly, performing convolution on the layer by using 64 convolution kernels with 1 × 1 to obtain a score _1 layer, and the characteristic size is 16 × 32 × 64; then, summing the score _1 layer and the deconv _1 layer to obtain a fuse _1 layer, wherein the characteristic size is still 16 × 32 × 64; finally, deconv operation is carried out on the fuse _1 layer by using 256 convolution kernels of 4 × 4 to obtain a deconv _2 layer, wherein the characteristic size is 32 × 64;
3. processing a resnet _ v1_50/block2/unit _ 1/botteleck _ v1 layer, wherein the characteristic size of the layer is 32 × 64 × 512, firstly, performing convolution on the layer by using 64 convolution kernels with 1 × 1 to obtain a score _2 layer, wherein the characteristic size is 32 × 64; then, summing the score _2 layer and the deconv _2 layer to obtain a fuse _2 layer, wherein the characteristic size is still 16 × 32 × 64; then, performing deconv operation on the fuse _2 layer by using 256 convolution kernels of 16 × 16 to obtain a deconv _ final layer, wherein the characteristic size of the deconv _ final layer is 256 × 512 × 64; finally, 2 convolution kernels of 1 × 1 are used for convolution operation on the deconv _ final layer to obtain a score _ final layer, and the corresponding feature size is 256 × 512 × 2.
The deconvolution deconv is a very common way to up-sample an image, and is commonly used in example segmentation, and its specific implementation process is essentially to obtain an image with a required spatial resolution by performing a convolution operation after filling an input.
In a second embodiment of the present invention, fig. 5(a) is a schematic diagram of an original picture of the second embodiment of the present invention; fig. 5(b) is a diagram illustrating the detection result of the second embodiment of the present invention. Fig. 5(a) and 5(b) are schematic diagrams of an original picture and a detection result thereof according to a second embodiment of the present invention, which show a detection result in a scene with a drastic change in illumination. Wherein, the detection result under different illumination conditions can be identified by different colors, for example: where the illumination is sufficient, the lane line may be yellow in color, and where the illumination is insufficient, the lane line may be purple in color, and so on. As can be seen from fig. 5(a) and 5(b), the detection result of the second embodiment of the present invention in a scene with a drastic change in illumination (as shown in fig. 5 (b)) is substantially the same as the original picture (as shown in fig. 5 (a)), and the accuracy of the detection result is high.
According to the technical solution of another embodiment of the present invention, the network of the bottom layer coding model Encoder may also be modified, for example, some simplified residual error networks, such as MobileNet, may be considered from the real-time point of view, and similarly, the decoding model Decoder portion still uses the full convolution neural network FCN. After a large amount of data training and result analysis, the best effect is obtained when the selected decoding layers are the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise layer of the point-state type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise layer of the point-state type and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise layer of the point-state type.
Fig. 6 is a schematic diagram of a process of decoding a decoding layer according to another embodiment of the present invention. As shown in fig. 6, the specific process of decoding these decoding layers to realize feature recognition is as follows:
1. processing the conv2d _12_ position layer, wherein the obtained characteristic size of the conv2d _12_ position layer is 8 × 16 × 1024, and firstly, using 256 convolution kernels of 7 × 7 to obtain a fully connected layer fc6(full volume) layer, wherein the characteristic size is 8 × 16 × 256; then 256 convolution kernels of 1 × 1 are used for the fc6 layer to obtain an fc7 (fusion convolution) layer with the characteristic size of 8 × 16 × 256; then, 64 convolution kernels of 1 × 1 were applied to fc7 layers to obtain score _ orignal layers with a characteristic size of 8 × 16 × 64; finally, carrying out deconvolution (deconv) operation on the score _ original layer by using 256 convolution kernels with 4 × 4 to obtain a deconv _1 layer with the characteristic size of 16 × 32 × 64;
2. processing a conv2d _6_ pointwise layer, wherein the characteristic size of the conv2d _6_ pointwise layer is 16 × 32 × 512, and firstly, convolving the conv2d _6_ pointwise layer by using 64 convolution kernels of 1 × 1 to obtain a score _1 layer, wherein the characteristic size is 16 × 32 × 64; then, summing score _1 layer and deconv _1 layer to obtain fuse _1 layer, wherein the characteristic size of the fuse _1 layer is still 16 × 32 × 64; finally, deconv operation is carried out on the fuse _1 layer by using 256 convolution kernels of 4 × 4 to obtain a deconv _2 layer, wherein the characteristic size is 32 × 64;
3. processing a conv2d _4_ pointwise layer, wherein the characteristic size of the conv2d _4_ pointwise layer is 32 × 64 × 256, and firstly, carrying out convolution operation on the conv2d _4_ pointwise layer by using 64 convolution kernels of 1 × 1 to obtain a score _2 layer, wherein the characteristic size is 32 × 64; then summing the score _2 layer and the deconv _2 layer to obtain a fuse _2 layer, wherein the characteristic size is 32 x 64; then, deconv operation is carried out on the fuse _2 layer by using 256 convolution kernels of 16 × 16 to obtain a deconv _ final layer, wherein the characteristic size is 256 × 512 × 64; finally, convolution operation is performed on the deconv _ final layer by using 2 convolution kernels of 1 × 1 to obtain score _ final with the size of 256 × 512 × 2.
FIG. 7(a) is a diagram of an original picture according to a third embodiment of the present invention; fig. 7(b) is a diagram illustrating the detection result of the third embodiment of the present invention. Fig. 7(a) and 7(b) are schematic diagrams of an original picture and a detection result thereof according to a third embodiment of the present invention, which show a detection result of a lane line based on a coding model MobileNet and a decoding model FCN. Wherein, the detection results of different lane lines can be marked by different colors. As can be seen from fig. 7(a) and 7(b), the detection result of the third embodiment of the present invention (shown in fig. 7 (b)) is substantially the same as the original picture (shown in fig. 7 (a)), and the detection result has high accuracy.
It can be seen from the above description that the method for detecting lane lines of the present invention obtains the binary bitmap and the example graph of the lane lines according to the original picture of the lane lines; then, the original picture, the binary bitmap and the example graph are input into the neural network for lane line detection according to the embodiment of the present invention to perform feature extraction and feature recognition, thereby performing lane line detection. When the binary bitmap and the example graph of the lane line are obtained according to the original picture of the lane line, processing can be performed through a cubic spline interpolation method, and therefore data input into the coding model are more accurate. It should be understood by those skilled in the art that, in the neural network for lane line detection according to the embodiment of the present invention, the residual error network used in the coding model is not limited to ResNet-50 and MobileNet, and other residual error networks are as follows: ResNet-101, ResNet-152, etc. are also functional, with only the decoding layers chosen for each residual network being different.
Fig. 8 is a schematic diagram of main blocks of an apparatus for lane line detection according to an embodiment of the present invention. As shown in fig. 8, the apparatus 800 for detecting lane lines according to the embodiment of the present invention mainly includes an image acquisition module 801, a feature extraction module 802, and a feature recognition module 803.
An image obtaining module 801, configured to obtain a binary bitmap and an example graph of a lane line according to an original picture of the lane line;
a feature extraction module 802, configured to perform feature extraction on the original picture, the binary bitmap, and the instance graph by using a residual error network;
a feature identification module 803, configured to perform feature identification on the extracted features using a full convolutional network to perform lane line detection.
According to an embodiment of the present invention, the image acquisition module 801 may further be configured to:
obtaining coordinates of a series of points on the lane line according to the original picture of the lane line;
performing cubic spline interpolation on the series of points to obtain a lane line trajectory line;
setting pixel values of points located on a track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line;
the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example graph of the lane line.
According to another embodiment of the present invention, the feature recognition module 803 may be further configured to:
selecting a decoding layer for feature recognition, wherein the decoding layer is a designated layer of the residual error network;
and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network so as to perform feature identification.
According to another embodiment of the present invention, the residual network is ResNet-50, and the selected decoding layers are: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
According to another embodiment of the present invention, the residual network is MobileNet, and the selected decoding layers are: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
According to the technical scheme of the embodiment of the invention, a binary bitmap and an example graph of the lane line are obtained according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; the extracted features are subjected to feature recognition by using a full convolution network so as to detect the lane lines, so that the problem of low accuracy of a detection result caused by the fact that a feature extraction bottom layer model is too simple is solved; the method can ensure the processing performance of the network model and improve the accuracy of lane line detection while using the complex network model to perform feature extraction and feature identification, so that higher accuracy can be obtained under complex scenes with larger changes such as illumination and the like.
Fig. 9 shows an exemplary system architecture 900 to which the method of lane line detection or the apparatus of lane line detection of the embodiments of the present invention may be applied.
As shown in fig. 9, the system architecture 900 may include end devices 901, 902, 903, a network 904, and a server 905. Network 904 is the medium used to provide communication links between terminal devices 901, 902, 903 and server 905. Network 904 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
A user may use the terminal devices 901, 902, 903 to interact with a server 905 over a network 904 to receive or send messages and the like. The terminal devices 901, 902, 903 may have installed thereon various messenger client applications such as, for example only, a shopping-like application, a web browser application, a search-like application, an instant messaging tool, a mailbox client, social platform software, etc.
The terminal devices 901, 902, 903 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.
The server 905 may be a server providing various services, such as a background management server (for example only) providing support for shopping websites browsed by users using the terminal devices 901, 902, 903. The backend management server may analyze and perform other processing on the received data such as the product information query request, and feed back a processing result (for example, target push information, product information — just an example) to the terminal device.
It should be noted that the method for detecting lane lines provided in the embodiment of the present invention is generally executed by the server 905, and accordingly, the apparatus for detecting lane lines is generally disposed in the server 905.
It should be understood that the number of terminal devices, networks, and servers in fig. 9 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
Referring now to FIG. 10, a block diagram of a computer system 1000 suitable for use with a terminal device or server implementing an embodiment of the invention is shown. The terminal device or the server shown in fig. 10 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.
As shown in fig. 10, the computer system 1000 includes a Central Processing Unit (CPU)1001 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)1002 or a program loaded from a storage section 1008 into a Random Access Memory (RAM) 1003. In the RAM 1003, various programs and data necessary for the operation of the system 1000 are also stored. The CPU 1001, ROM 1002, and RAM 1003 are connected to each other via a bus 1004. An input/output (I/O) interface 1005 is also connected to bus 1004.
The following components are connected to the I/O interface 1005: an input section 1006 including a keyboard, a mouse, and the like; an output section 1007 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage portion 1008 including a hard disk and the like; and a communication section 1009 including a network interface card such as a LAN card, a modem, or the like. The communication section 1009 performs communication processing via a network such as the internet. The driver 1010 is also connected to the I/O interface 1005 as necessary. A removable medium 1011 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 1010 as necessary, so that a computer program read out therefrom is mounted into the storage section 1008 as necessary.
In particular, according to the embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication part 1009 and/or installed from the removable medium 1011. The computer program executes the above-described functions defined in the system of the present invention when executed by the Central Processing Unit (CPU) 1001.
It should be noted that the computer readable medium shown in the present invention can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present invention, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present invention, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units or modules described in the embodiments of the present invention may be implemented by software, or may be implemented by hardware. The described units or modules may also be provided in a processor, and may be described as: a processor includes an image acquisition module, a feature extraction module, and a feature recognition module. Where the names of these units or modules do not in some cases constitute a limitation of the units or modules themselves, for example, the image acquisition module may also be described as a "module for deriving a binary bitmap and an instance map of a lane line from an original picture of the lane line".
As another aspect, the present invention also provides a computer-readable medium that may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise: obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; feature recognition is performed on the extracted features using a full convolutional network for lane line detection.
According to the technical scheme of the embodiment of the invention, a binary bitmap and an example graph of the lane line are obtained according to the original picture of the lane line; extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network; the extracted features are subjected to feature recognition by using a full convolution network so as to detect the lane lines, so that the problem of low accuracy of a detection result caused by the fact that a feature extraction bottom layer model is too simple is solved; the method can ensure the processing performance of the network model and improve the accuracy of lane line detection while using the complex network model to perform feature extraction and feature identification, so that higher accuracy can be obtained under complex scenes with larger changes such as illumination and the like.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (12)
1. A method of lane line detection, comprising:
obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line;
extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network;
feature recognition is performed on the extracted features using a full convolutional network for lane line detection.
2. The method of claim 1, wherein obtaining the binary bitmap and the instance graph of the lane line from the original picture of the lane line comprises:
obtaining coordinates of a series of points on the lane line according to the original picture of the lane line;
performing cubic spline interpolation on the series of points to obtain a lane line trajectory line;
setting pixel values of points located on a track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line;
the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example graph of the lane line.
3. The method of claim 1, wherein feature identifying the extracted features using a full convolutional network comprises:
selecting a decoding layer for feature recognition, wherein the decoding layer is a designated layer of the residual error network;
and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network so as to perform feature identification.
4. The method of claim 3 wherein the residual network is ResNet-50 and the decoding layers selected are: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
5. The method of claim 3 wherein the residual network is a MobileNet and the selected decoding layers are: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
6. An apparatus for lane line detection, comprising:
the image acquisition module is used for obtaining a binary bitmap and an example graph of the lane line according to the original picture of the lane line;
the characteristic extraction module is used for extracting the characteristics of the original picture, the binary bitmap and the example graph by using a residual error network;
and the characteristic identification module is used for carrying out characteristic identification on the extracted characteristics by using a full convolution network so as to detect the lane line.
7. The apparatus of claim 6, wherein the image acquisition module is further configured to:
obtaining coordinates of a series of points on the lane line according to the original picture of the lane line;
performing cubic spline interpolation on the series of points to obtain a lane line trajectory line;
setting pixel values of points located on a track line trajectory line to [255, 255, 255], and pixel values of points located at the rest positions to [0, 0, 0] to obtain a binary bitmap of the track line;
the pixel values of the points located on the different lane line trajectory lines are respectively set to different levels other than 0, and the pixel values of the points located at the remaining positions are set to [0, 0, 0] to obtain an example graph of the lane line.
8. The apparatus of claim 6, wherein the feature identification module is further configured to:
selecting a decoding layer for feature recognition, wherein the decoding layer is a designated layer of the residual error network;
and performing convolution operation on the extracted features corresponding to the decoding layer by using a full convolution network so as to perform feature identification.
9. The apparatus of claim 8 wherein the residual network is ResNet-50 and the decoding layers selected are: the resnet _ V1_50/block4 layer represented by the layer resnet _ V1_50/block4 represented by the 4 th module of the resnet network of the V1-50 version, the resnet _ V1_50/block3/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 3 rd module of the resnet network of the V1-50 version, and the resnet _ V1_50/block2/unit _ 1/cottleneck _ V1 layer represented by the bottleneck structure of the V1 version of the 1 st unit of the 2 nd module of the resnet network of the V1-50 version.
10. The apparatus of claim 8 wherein the residual network is MobileNet and the selected decoding layers are: the 12 th 2-dimensional convolutional layer conv2d _12_ pointwise of the dot type, the 6 th 2-dimensional convolutional layer conv2d _6_ pointwise of the dot type, and the 4 th 2-dimensional convolutional layer conv2d _4_ pointwise of the dot type.
11. An electronic device that performs lane line detection, comprising:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of claims 1-5.
12. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910990018.3A CN112686080A (en) | 2019-10-17 | 2019-10-17 | Method and device for detecting lane line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910990018.3A CN112686080A (en) | 2019-10-17 | 2019-10-17 | Method and device for detecting lane line |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112686080A true CN112686080A (en) | 2021-04-20 |
Family
ID=75444744
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910990018.3A Pending CN112686080A (en) | 2019-10-17 | 2019-10-17 | Method and device for detecting lane line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686080A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778668A (en) * | 2016-12-30 | 2017-05-31 | 明见（厦门）技术有限公司 | A kind of method for detecting lane lines of the robust of joint RANSAC and CNN |
| US20180181817A1 (en) * | 2015-09-10 | 2018-06-28 | Baidu Online Network Technology (Beijing) Co., Ltd. | Vehicular lane line data processing method, apparatus, storage medium, and device |
| US20180225527A1 (en) * | 2015-08-03 | 2018-08-09 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method, apparatus, storage medium and device for modeling lane line identification, and method, apparatus, storage medium and device for identifying lane line |
| US20180373981A1 (en) * | 2017-06-21 | 2018-12-27 | TuSimple | Method and device for optimizing neural network |
| CN109635744A (en) * | 2018-12-13 | 2019-04-16 | 合肥工业大学 | A kind of method for detecting lane lines based on depth segmentation network |
| CN109886176A (en) * | 2019-02-14 | 2019-06-14 | 武汉大学 | Method for detecting lane lines under complicated Driving Scene |
| CN110197151A (en) * | 2019-05-28 | 2019-09-03 | 大连理工大学 | A kind of lane detection system and method for combination double branching networks and custom function network |
-
2019
- 2019-10-17 CN CN201910990018.3A patent/CN112686080A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180225527A1 (en) * | 2015-08-03 | 2018-08-09 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method, apparatus, storage medium and device for modeling lane line identification, and method, apparatus, storage medium and device for identifying lane line |
| US20180181817A1 (en) * | 2015-09-10 | 2018-06-28 | Baidu Online Network Technology (Beijing) Co., Ltd. | Vehicular lane line data processing method, apparatus, storage medium, and device |
| CN106778668A (en) * | 2016-12-30 | 2017-05-31 | 明见（厦门）技术有限公司 | A kind of method for detecting lane lines of the robust of joint RANSAC and CNN |
| US20180373981A1 (en) * | 2017-06-21 | 2018-12-27 | TuSimple | Method and device for optimizing neural network |
| CN109635744A (en) * | 2018-12-13 | 2019-04-16 | 合肥工业大学 | A kind of method for detecting lane lines based on depth segmentation network |
| CN109886176A (en) * | 2019-02-14 | 2019-06-14 | 武汉大学 | Method for detecting lane lines under complicated Driving Scene |
| CN110197151A (en) * | 2019-05-28 | 2019-09-03 | 大连理工大学 | A kind of lane detection system and method for combination double branching networks and custom function network |
Non-Patent Citations (4)
| Title |
|---|
| KIM J 等: "Robust lane detection based on convolutional neural network and random sample consensus", PROCEEDINGS OF THE21ST INTERNATIONAL CONFERENCE ON NEURAL INFORMATION PROCESSING * |
| 刘宇: "基于卷积神经网络的道路目标检测算法研究与实现", 中国优秀硕士学位论文全文数据库信息科技辑, vol. 2018, no. 12, pages 3 * |
| 刘波;郭申;: "基于卷积定理的人脸验证CNN模型加速", 北京工业大学学报, no. 11 * |
| 贾瑞明 等: "基于编解码卷积神经网络的单张图像深度估计", 《图学学报》, vol. 40, no. 4, pages 3 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10650236B2 (en) | Road detecting method and apparatus | |
| US11151723B2 (en) | Image segmentation method, apparatus, and fully convolutional network system | |
| CN107944450B (en) | License plate recognition method and device | |
| US11216924B2 (en) | Method and apparatus for processing image | |
| CN112184738B (en) | Image segmentation method, device, equipment and storage medium | |
| EP3815047A1 (en) | Image colorization based on reference information | |
| CN110288549B (en) | Video repairing method and device and electronic equipment | |
| WO2021115061A1 (en) | Image segmentation method and apparatus, and server | |
| CN112016510A (en) | Signal lamp identification method and device based on deep learning, equipment and storage medium | |
| CN110598705B (en) | Semantic annotation method and device for image | |
| CN114140592A (en) | High-precision map generation method, device, equipment, medium and automatic driving vehicle | |
| CN113052143A (en) | Handwritten digit generation method and device | |
| CN113421217A (en) | Method and device for detecting travelable area | |
| CN110633598B (en) | Method and device for determining a driving area in an environment image | |
| CN112686080A (en) | Method and device for detecting lane line | |
| CN113411550B (en) | Video coloring method, device, equipment and storage medium | |
| CN110634155A (en) | Target detection method and device based on deep learning | |
| CN114419070A (en) | Image scene segmentation method, device, equipment and storage medium | |
| CN114494782A (en) | Image processing method, model training method, related device and electronic equipment | |
| CN113947529A (en) | Image enhancement method, model training method, component identification method and related equipment | |
| CN112967191A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN112132031A (en) | Vehicle money identification method and device, electronic equipment and storage medium | |
| CN112434591B (en) | Lane line determination method and device | |
| CN115049895B (en) | Image attribute identification method, attribute identification model training method and device | |
| CN113362421B (en) | Drawing method and device of diversion area in map and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |