CN113537147A - Night lane line detection method based on illumination compensation - Google Patents
Night lane line detection method based on illumination compensation Download PDFInfo
- Publication number
- CN113537147A CN113537147A CN202110906427.8A CN202110906427A CN113537147A CN 113537147 A CN113537147 A CN 113537147A CN 202110906427 A CN202110906427 A CN 202110906427A CN 113537147 A CN113537147 A CN 113537147A
- Authority
- CN
- China
- Prior art keywords
- lane line
- clustering
- image
- slope
- lane
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a night lane line detection method based on illumination compensation, which is characterized by comprising the following steps of: 1) obtaining an intrinsic mode component image IMG1 and an illumination compensated image IMG 2; 2) fusing IMG1 and IMG 2; 3) denoising the enhanced image by Gaussian filtering, carrying out edge detection by using a canny edge detection algorithm, and carrying out line detection by using Hough transformation; 4) and classifying and clustering the lane lines step by utilizing the slope and intercept characteristics of the lane lines, eliminating false lane lines, and finally fitting the lane line segments reserved after clustering by utilizing a least square method to obtain the final lane lines. The method can enhance the brightness and contrast of the night image, and can remove the non-lane line noise generated by complex illumination, and the method can realize good detection of the lane line under various weather and illumination conditions at night.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a night lane line detection method based on illumination compensation.
Background
In recent years, attention has been paid to and Advanced in the art of intelligent vehicle safety represented by Advanced Driver Assistance Systems (ADAS). A key technology in ADAS, namely, a Lane departure warning system LDWS (short for LDWS), is based on Lane line detection.
At present, most of the research on lane line detection is based on the conventional illumination condition (such as daytime), the definition of an image acquired in the environment is high, and the contrast between a lane line and a background road surface is obvious. However, the night illumination is uneven, the contrast is low, the overall brightness of the image is low, and the effect of the lane line detection method under the conventional illumination condition is sharply reduced under the condition, so that the requirement of the lane line detection at night cannot be met. To address this problem, Yoo et al use maximizing the gradient between the lane line and the road surface in the image to enhance the contrast between the lane line and the road surface; mei et al enhance low contrast areas using histogram equalization; jongin et al use the color characteristics of lane lines under different lighting conditions to reduce the effect of lighting on the lane lines; wang et al propose a method of nonlinear contrast stretching to enhance images; roche and the like take color and illumination information as image brightness characteristics and design a classifier, and train and test the image. The method can obtain ideal effect under the condition of relatively good image illumination, but has poor detection effect on the complex night environment such as weak road light or no street lamp and the like.
Disclosure of Invention
The invention aims to provide a night lane line detection method based on illumination compensation aiming at the defects of the prior art. The method can enhance the brightness and contrast of the night image, and can remove the non-lane line noise generated by complex illumination, and the method can realize good detection of the lane line under various weather and illumination conditions at night.
The technical scheme for realizing the purpose of the invention is as follows:
a night lane line detection method based on illumination compensation comprises the following steps:
carrying out gray level transformation and denoising processing on the obtained image;
1) firstly, carrying out gray level transformation and denoising on an obtained image, and then respectively adopting empirical mode decomposition and a multi-scale retinex algorithm to the denoised gray level image to obtain an intrinsic mode component image IMG1 and an illumination compensated image IMG 2;
2) fusing the intrinsic mode component image IMG1 and the illumination compensated image IMG2 to obtain an enhanced image, namely fusing the IMG1 and the IMG2 as follows:
IMG = a × IMG1+ IMG2, IMG being the enhanced image, a being a coefficient factor greater than 1;
3) and denoising the enhanced image by Gaussian filtering, carrying out edge detection by using a canny edge detection algorithm, and carrying out line detection by using Hough transformation. Removing edges formed by night noise and pavement cracks by adopting Gaussian filtering with the Gaussian kernel size of 5 x 5, wherein a canny edge detection threshold value is [80,170 ];
4) dividing the detected straight line into a left lane line and a right lane line by adopting lane line characteristics, respectively clustering the left lane line and the right lane line step by step, namely dividing the lane lines into the left lane line and the right lane line according to the condition that the slope of the left lane line is less than zero and the slope of the right lane line is greater than zero, pre-screening the lane lines according to the condition that the slope of the lane lines is between 25 degrees and 65 degrees and between 115 degrees and 155 degrees, clustering the left lane line and the right lane line step by step according to the slope and intercept characteristics of the left lane line and the right lane line, and fitting the lane lines by adopting a least square method according to the clustered results.
The step-by-step clustering in the step 4) comprises the following steps:
1-4) Clustering the slopes of the left and right lane lines by adopting a Density-Based Spatial Clustering method DBSCAN (DBSCAN), selecting the class with the largest number in the slope Clustering results as a first Clustering result, and obtaining the lane lines with similar slopes, wherein the slope Clustering radius is 0.5;
2-4) clustering the left/right lane lines with similar slope according to intercept characteristics to obtain the lane lines with similar slope and intercept as a final lane line screening result, wherein the intercept clustering radius is determined according to the following formula:
Δ b and Δ k represent the intercept difference and the slope difference of two straight lines, respectively, and W represents the width of the image.
According to the technical scheme, the MSR illumination compensation and the EMD image decomposition are firstly carried out on the gray level image respectively, in order to make up for the defects that the MSR illumination compensation image is fuzzy in edge and low in local contrast, the illumination compensation image and the built-in mode component image generated by EMD decomposition are fused, and the enhanced image with good illumination and clear edge is obtained. In order to eliminate false lane lines generated by the enhanced image due to complex illumination, a spatial clustering method based on density is utilized to cluster the lane lines step by step.
According to the technical scheme, the multi-scale retinex algorithm is used for illumination compensation of the gray level image at night, so that the overall brightness of the image is improved; decomposing the night gray level image by using an empirical mode to obtain an intrinsic mode component image generated by first decomposition, and overlapping the intrinsic mode component image and an image subjected to multi-scale retinex illumination compensation, so that the defects of blurred edge and insufficient local contrast of the illumination compensation image are overcome; and the left lane line and the right lane line are clustered step by adopting a spatial clustering method based on density, so that false lane lines generated by the enhanced image due to complex illumination are eliminated.
The method can enhance the brightness and contrast of the night image, and can remove the non-lane line noise generated by complex illumination, and the method can realize good detection of the lane line under various weather and illumination conditions at night.
Drawings
FIG. 1 is a schematic flow chart of an exemplary method;
FIG. 2 is an image after illumination compensation in an embodiment;
FIG. 3 is an IMF image obtained by a first decomposition of an empirical mode decomposition in an embodiment;
FIG. 4 is an image after enhancement in the embodiment;
FIG. 5 is an image of the left and right lane lines screened and divided in the example;
FIG. 6 shows the result of the lane line clustering in the example;
fig. 7 shows the final lane line detection result in the example.
Detailed Description
The present invention will be described in further detail with reference to the following drawings and examples, but the present invention is not limited thereto.
Example (b):
referring to fig. 1, a method for detecting a lane line at night based on illumination compensation includes the steps of:
1) firstly, gray level transformation and denoising processing are carried out on an obtained image, and then an intrinsic mode component image IMG1 and an illumination compensated image IMG2 are obtained on the denoised gray level image by adopting empirical mode decomposition and a multi-scale retinex algorithm respectively, as shown in FIGS. 2 and 3;
2) fusing the intrinsic mode component image IMG1 and the illumination compensated image IMG2 to obtain an enhanced image, namely fusing the IMG1 and the IMG2 as follows:
IMG = a, IMG1+ IMG2, where IMG is an enhanced image, and as shown in fig. 4, because the illumination at night is weak, the luminance of an IMF image generated by empirical mode decomposition is small, and the IMF image is multiplied by a coefficient factor a greater than 1 during fusion to improve the edge luminance of the fused image;
3) the lane line image obtained by the camera with uneven illumination at night and low brightness contains larger noise compared with the daytime, the noise is represented as small edges in the enhanced image, meanwhile, the obvious edges can be generated due to the difference of cracks and chromaticity of the road surface, the edges can be removed through small-scale Gaussian blur compared with the lane line, and the enhanced image is subjected to Gaussian filtering denoising, canny edge detection algorithm edge detection and Hough transformation line detection. Wherein, the edges formed by night noise and pavement cracks are removed by adopting Gaussian filtering with the Gaussian kernel size of 5 x 5, the canny edge detection threshold value is [80,170], and the shortest line length is set to be 30 pixels in Hough linear detection;
4) dividing the detected straight line into a left lane line and a right lane line by adopting lane line characteristics, respectively clustering the left lane line and the right lane line step by step, namely dividing the lane lines into the left lane line and the right lane line according to the condition that the slope of the left lane line is less than zero and the slope of the right lane line is greater than zero, pre-screening the lane lines according to the condition that the slope of the lane lines is between 25 degrees and 65 degrees and between 115 degrees and 155 degrees, and respectively clustering the left lane line and the right lane line step by step according to the slope and intercept characteristics of the left lane line and the right lane line, as shown in figure 5, fitting the clustered lane lines into straight lines by adopting a least square method to obtain a final lane line detection result, as shown in figure 7.
The step-by-step clustering in the step 4) comprises the following steps:
1-4) clustering the slopes of the left/right lane lines by adopting a density-based spatial clustering method DBSCAN, selecting the class with the largest number in the slope clustering results as a first clustering result, and obtaining the lane lines with similar slopes, wherein the slope clustering radius is 0.5;
2-4) clustering the left/right lane lines with similar slope according to intercept characteristics to obtain the lane lines with similar slope and intercept as a final lane line screening result, wherein as shown in FIG. 6, the intercept clustering radius is determined according to the following formula:
Δ b and Δ k represent the intercept difference and the slope difference of two straight lines, respectively, and W represents the width of the image.
Claims (2)
1. A night lane line detection method based on illumination compensation is characterized by comprising the following steps:
1) respectively adopting empirical mode decomposition and a multi-scale retinex algorithm to obtain an intrinsic mode component image IMG1 and an illumination compensated image IMG2 for the gray level image;
2) fusing the intrinsic mode component image IMG1 and the illumination compensated image IMG2 to obtain an enhanced image, namely fusing the IMG1 and the IMG2 as follows:
IMG = a × IMG1+ IMG2, IMG being the enhanced image, a being a coefficient factor greater than 1;
3) removing noise of the enhanced image by Gaussian filtering, performing edge detection by using a canny edge detection algorithm, and performing line detection by using Hough transformation, wherein edges formed by night noise and pavement cracks are removed by adopting Gaussian filtering with a Gaussian kernel size of 5 x 5, and a canny edge detection threshold value is [80,170 ];
4) dividing the detected straight line into a left lane line and a right lane line by adopting lane line characteristics, and respectively clustering the left lane line and the right lane line step by step, namely dividing the lane lines into a left lane line and a right lane line according to the condition that the slope of the left lane line is less than zero and the slope of the right lane line is greater than zero, pre-screening the lane lines according to the condition that the slope of the lane lines is between 25 degrees and 65 degrees and 115 degrees and 155 degrees, and respectively clustering the left lane line and the right lane line step by step according to the slope and intercept characteristics of the left lane line and the right lane line.
2. The illumination compensation-based nighttime lane line detection method according to claim 1, wherein the step-by-step clustering in step 4) comprises:
1-4) clustering the slopes of the left/right lane lines by adopting a density-based spatial clustering method DBSCAN, selecting the class with the largest number in the slope clustering results as a first clustering result, and obtaining the lane lines with similar slopes, wherein the slope clustering radius is 0.5;
2-4) clustering the left/right lane lines with similar slope according to slope characteristics to obtain the lane lines with similar slope and intercept as a final lane line screening result, wherein the intercept clustering radius is determined according to the following formula:
Δ b and Δ k represent the intercept difference and the slope difference of two straight lines, respectively, and W represents the width of the image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110906427.8A CN113537147B (en) | 2021-08-09 | 2021-08-09 | Night lane line detection method based on illumination compensation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110906427.8A CN113537147B (en) | 2021-08-09 | 2021-08-09 | Night lane line detection method based on illumination compensation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113537147A true CN113537147A (en) | 2021-10-22 |
CN113537147B CN113537147B (en) | 2022-04-12 |
Family
ID=78122156
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110906427.8A Active CN113537147B (en) | 2021-08-09 | 2021-08-09 | Night lane line detection method based on illumination compensation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113537147B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114387582A (en) * | 2022-01-13 | 2022-04-22 | 福州大学 | Lane detection method under poor illumination condition |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810677A (en) * | 2014-01-16 | 2014-05-21 | 河南科技大学 | Remote sensing image enhancement method based on super-empirical mode decomposition |
US9305362B1 (en) * | 2014-02-28 | 2016-04-05 | Xilinx, Inc. | Image stabilization |
US20180253830A1 (en) * | 2017-02-24 | 2018-09-06 | Sunnybrook Research Institute | Systems and methods for noise reduction in imaging |
CN108875657A (en) * | 2018-06-26 | 2018-11-23 | 北京茵沃汽车科技有限公司 | A kind of method for detecting lane lines |
CN109409355A (en) * | 2018-08-13 | 2019-03-01 | 国网陕西省电力公司 | A kind of method and device of novel transformer nameplate identification |
CN110909686A (en) * | 2019-11-26 | 2020-03-24 | 黑龙江大学 | Low-illumination image enhancement system for driving assistance |
CN111079598A (en) * | 2019-12-06 | 2020-04-28 | 深圳市艾为智能有限公司 | Lane line detection method based on image texture and machine learning |
CN112529049A (en) * | 2020-11-24 | 2021-03-19 | 中国科学院植物研究所 | Urban green land species shade tolerance evaluation method and urban green land species selection method |
-
2021
- 2021-08-09 CN CN202110906427.8A patent/CN113537147B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103810677A (en) * | 2014-01-16 | 2014-05-21 | 河南科技大学 | Remote sensing image enhancement method based on super-empirical mode decomposition |
US9305362B1 (en) * | 2014-02-28 | 2016-04-05 | Xilinx, Inc. | Image stabilization |
US20180253830A1 (en) * | 2017-02-24 | 2018-09-06 | Sunnybrook Research Institute | Systems and methods for noise reduction in imaging |
CN108875657A (en) * | 2018-06-26 | 2018-11-23 | 北京茵沃汽车科技有限公司 | A kind of method for detecting lane lines |
CN109409355A (en) * | 2018-08-13 | 2019-03-01 | 国网陕西省电力公司 | A kind of method and device of novel transformer nameplate identification |
CN110909686A (en) * | 2019-11-26 | 2020-03-24 | 黑龙江大学 | Low-illumination image enhancement system for driving assistance |
CN111079598A (en) * | 2019-12-06 | 2020-04-28 | 深圳市艾为智能有限公司 | Lane line detection method based on image texture and machine learning |
CN112529049A (en) * | 2020-11-24 | 2021-03-19 | 中国科学院植物研究所 | Urban green land species shade tolerance evaluation method and urban green land species selection method |
Non-Patent Citations (2)
Title |
---|
FARZIN MATIN: ""Retinex-Based Image Enhancement with Particle Swarm Optimization and Multi-Objective Function"", 《IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS》 * |
李超: ""一种基于帧间关联的实时车道线检测算法"", 《计算机科学》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114387582A (en) * | 2022-01-13 | 2022-04-22 | 福州大学 | Lane detection method under poor illumination condition |
Also Published As
Publication number | Publication date |
---|---|
CN113537147B (en) | 2022-04-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
TWI607901B (en) | Image inpainting system area and method using the same | |
CN105981042B (en) | Vehicle detection system and method | |
CN104050450A (en) | Vehicle license plate recognition method based on video | |
CN110544211B (en) | Method, system, terminal and storage medium for detecting lens attached object | |
CN109886168B (en) | Ground traffic sign identification method based on hierarchy | |
CN102938057A (en) | Vehicle shadow eliminating method and device | |
CN104616258A (en) | Rapid defogging method for road image | |
CN104598928A (en) | Vehicle classification method based on class dictionary sparse identification | |
CN107133588A (en) | Vehicle identification method based on vehicle window feature extraction | |
CN111598814B (en) | Single image defogging method based on extreme scattering channel | |
CN110276318A (en) | Nighttime road rains recognition methods, device, computer equipment and storage medium | |
CN113537147B (en) | Night lane line detection method based on illumination compensation | |
Mandal et al. | Real-time automotive night-vision system for drivers to inhibit headlight glare of the oncoming vehicles and enhance road visibility | |
CN112464731A (en) | Traffic sign detection and identification method based on image processing | |
Sultana et al. | Vision-based robust lane detection and tracking in challenging conditions | |
CN107437241B (en) | Dark channel image defogging method combined with edge detection | |
CN110853000B (en) | Rut detection method | |
KR20130070210A (en) | Method for removing noise of image | |
CN112308801A (en) | Road traffic tracking management method and system based on big data image acquisition | |
Nguyen et al. | Fast traffic sign detection under challenging conditions | |
Khoshki et al. | Improved Automatic License Plate Recognition (ALPR) system based on single pass Connected Component Labeling (CCL) and reign property function | |
Ukani et al. | An accurate method for license plate localization using morphological operations and edge processing | |
CN113743226B (en) | Daytime front car light language recognition and early warning method and system | |
CN115953341A (en) | Night rocket visible light image and infrared image regional fusion method | |
Pillai et al. | Night time vehicle detection using tail lights: a survey |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |