A lane line image region segmentation method

Abstract

The present invention relates to the field of computer vision technology, and particularly relates to a method for segmenting the lane line image area. The present invention processes the road image obtained by the vehicle side camera to achieve precise segmentation of the lane line area in the road image under complex scenarios, thereby improving the robustness of the lane line detection technology and the accuracy of vehicle positioning.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and particularly relates to a method for segmenting a lane line image area. Background Art

[0002] In the field of computer vision, the vehicle positioning technology based on lane line detection has become an important data support and effectiveness verification means for related algorithms such as vehicle lane change warning and lane departure warning. At present, this technology is mostly realized based on a front-view vehicle side camera installed at the vehicle rearview mirror. However, the vehicle side camera cannot directly obtain the image features between the current position of the vehicle and the lane line, and a complex image-spatial coordinate conversion algorithm is still required to complete the detection of the lane line. However, such algorithms are difficult to adapt to complex lighting scenarios (uneven brightness), lane line wear, and lane line interruption scenarios, and have poor robustness when processing real-time images in these scenarios, and are prone to problems of inaccurate vehicle positioning. Therefore, an image processing method is needed to process the images collected by the vehicle side camera in various scenarios to obtain a lane line area image more suitable for vehicle positioning algorithms. Summary of the Invention

[0003] Aiming at the problems existing in the prior art, the purpose of the present invention is to provide a method for fusing and segmenting a lane line image area.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions to be implemented.

[0005] A method for segmenting a lane line image area includes the following steps:

[0006] Step 1, preprocess the road image collected by the vehicle side camera;

[0007] Step 2, perform compensation processing on the preprocessed road image based on the illumination characteristics;

[0008] Step 3, respectively segment the lane line area of the compensated road image;

[0009] Step 4, select the best binary image of the lane line area.

[0010] Compared with the prior art, the beneficial effect of the present invention is: processing the road image obtained by the vehicle side camera to achieve precise segmentation of the lane line area in the road image under complex scenarios, thereby improving the robustness of the lane line detection technology and the accuracy of vehicle positioning. Brief Description of the Drawings

[0011] The following further describes the present invention in detail with reference to the drawings and specific embodiments.

[0012] Figure 1Schematic flowchart of the method of the present invention;

[0013] Figure 2 Example diagram of defining the region of interest (ROI) and adding a light compensation spline in the method of the present invention under three brightness scenarios; Figure 2 (a) In the near - dark and far - bright scenario; Figure 2 (b) In the uniform brightness scenario; Figure 2 (c) In the near - bright and far - dark scenario;

[0014] Figure 3 Image of a raw image captured by a side - mounted vehicle camera after defining the region of interest (ROI) and adding a light compensation spline;

[0015] Figure 4 is Figure 3 The grayscale histogram of sub - region B in the region of interest (ROI) without adding a light compensation spline;

[0016] Figure 5 is Figure 3 The grayscale histogram of sub - region B in the region of interest (ROI) after adding a light compensation spline;

[0017] Figure 6 Image of another raw image captured by a side - mounted vehicle camera after defining the region of interest (ROI) and adding a light compensation spline;

[0018] Figure 7 is Figure 6 The grayscale histogram of sub - region A in the region of interest (ROI) without adding a light compensation spline;

[0019] Figure 8 is Figure 6 The grayscale histogram of sub - region A in the region of interest (ROI) after adding a light compensation spline. Detailed implementation manners

[0020] The following will describe the implementation scheme of the present invention in detail in combination with embodiments. However, those skilled in the art will understand that the following embodiments are only used to illustrate the present invention and should not be regarded as limiting the scope of the present invention.

[0021] Referring to Figure 1 , a lane line image region segmentation method includes the following steps:

[0022] Step 1, pre - process the road image captured by the side - mounted vehicle camera;

[0023] Specifically, first, define the range of the Region of Interest (ROI) to exclude environmental interference factors such as the sky background and the road ahead; second, perform grayscale processing on the ROI image to eliminate redundant color information; finally, use the median filtering algorithm in the form of a sliding window to balance the gray value of the pixel at the center of the window and the gray values of the points in the window neighborhood, and eliminate the interference noise generated during the image acquisition process to complete the image preprocessing.

[0024] Step 2: Perform compensation processing on the preprocessed road image based on the illumination characteristics.

[0025] Specifically, first, divide the ROI range of the preprocessed road image longitudinally along the road into 4 sub-regions, as Figure 2 shown. The yellow thin lines in the figure are the boundaries of each sub-region. To minimize the differences in illumination non-uniformity and the existence of lane lines in the sub-regions; then, add three types of illumination compensation splines to the preprocessed road image respectively, and finally obtain three road images with different compensation characteristics.

[0026] Among them, the three types of illumination compensation splines are the illumination compensation spline for the near-dark and far-bright scene, the illumination compensation spline for the uniform brightness scene, and the illumination compensation spline for the near-bright and far-dark scene. The illumination compensation splines for each brightness scene have different brightness in each sub-region. The added illumination compensation splines are like the vertical slender splines on the right side of the ROI in Figure 2 .

[0027] Reference Figure 3 , the illumination brightness in the ROI is severely non-uniform. By adding the compensation spline, the average brightness of the peaks corresponding to the lane lines in the gray histogram of each sub-region in the ROI can be increased respectively, as Figure 4 , Figure 5 shown, to increase the brightness difference between the lane line area and the road surface area.

[0028] Reference Figure 6 , there is a break in the lane line in sub-region A of the ROI. By adding the compensation spline, a bimodal feature can be constructed for the gray histogram of the image in the lane line break area, as Figure 7 , Figure 8 shown.

[0029] Step 3: Perform lane line area segmentation on the compensated road images respectively.

[0030] Specifically, use the Otsu algorithm to perform lane line area segmentation on each sub-region of the compensated road image respectively, and finally obtain the binary image of the lane line area corresponding to the compensated road image; process the three road images with different compensation characteristics respectively, and finally obtain three binary images of the lane line area.

[0031] Considering the unevenness of light and the difference in the integrity of lane lines, the degree of lane line wear in different sub-areas is different. When the difference is large, if the Otsu algorithm is used to segment in the traditional way, the segmentation threshold of the severely worn area is bound to be too large, and the lane line area will be largely eliminated after segmentation. At this time, using the Otsu algorithm in each sub-area separately will avoid the above defects and preserve the severely worn lane line area. Therefore, completing the damaged lane line corresponds to the traditional use of the Otsu algorithm.

[0032] Step 4: Select the best lane line area binary image.

[0033] Specifically, firstly, morphological opening operations are performed on all lane line area binary images to eliminate lane line edge burrs; secondly, the cv2.findContours function built into Opencv is used to retrieve the number of white pixel set contours in the lane line area binary image; finally, the lane line area binary image with the number of white contours greater than 2 and closest to 2 is taken as the best lane line area binary image.

[0034] The best lane line area binary image is used as the input image of the vehicle positioning technology based on lane line detection, thereby improving the robustness of the lane line detection technology and the accuracy of vehicle positioning.

[0035] Although the present invention has been described in detail in general terms and in specific embodiments in this specification, it is obvious to those skilled in the art that some modifications or improvements may be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of protection claimed by the present invention.

Claims

1. A lane line image region segmentation method, characterized in that It includes the following steps: Step 1: Preprocess the road images collected by the vehicle side camera; Step 2: First, divide the preprocessed road image longitudinally along the road into 4 sub-regions; Then, add three types of light compensation splines to the preprocessed road image respectively, and finally obtain three road images with different compensation characteristics. Among them, the three types of light compensation splines are the light compensation spline for the near-dark and far-bright scene, the light compensation spline for the uniform brightness scene, and the light compensation spline for the near-bright and far-dark scene. The light compensation splines for each brightness scene have different brightness in each sub-region; Step 3: Use the Otsu algorithm to segment the lane line regions for each sub-region of the compensated road image respectively, and finally obtain the binary image of the lane line region corresponding to the compensated road image. Process the three road images with different compensation characteristics respectively, and finally obtain three binary images of the lane line regions; Step 4: Select the best binary image of the lane line region.

2. The lane line image region segmentation method according to claim 1, wherein Specifically for Step 1, first delimit the range of the region of interest (ROI); secondly, perform grayscale processing on the ROI image; finally, use the median filtering algorithm in the form of a sliding window to balance the gray value of the pixel at the center of the window and the gray values of the points in the window neighborhood.

3. The lane line image region segmentation method according to claim 1, characterized in that Specifically for Step 4, first, perform morphological opening operation on all binary images of the lane line regions respectively; secondly, retrieve the number of white pixel set contours in the binary image of the lane line region; finally, take the binary image of the lane line region with the number of white contours greater than 2 and closest to 2 as the best binary image of the lane line region.

4. The lane line image region segmentation method according to claim 3, wherein To retrieve the number of white pixel set contours in the binary image of the lane line region, specifically, use the cv2.findContours function built in Opencv for retrieval.

Images