SCIENTIA SINICA Informationis, Volume 47 , Issue 4 : 455-467(2017) https://doi.org/10.1360/N112016-00174

## Lane detection algorithm based on geometric moment sampling

• AcceptedOct 8, 2016
• PublishedFeb 10, 2017
Share
Rating

### Abstract

Existing lane detection algorithms come with costs regarding the required memory and the complexity of the algorithms limiting ways of generating versions adaptive to challenging road environments and portable low-power consumption equipments. We propose a new solution for lane detection based on geometric moment sampling. First, the current frame of a recorded video is processed by selecting potential regions of interest (i.e., potentially containing lane markings) based on piecewise sampling. Then, the visible road surface is segmented by labeling connected components of equivalent pixel values using an image binarization procedure. By analyzing the different order of geometric moments of connected components in the lane region, the centroid and the direction angle of detected parts of lane marking segments are calculated. We combine these calculated parameters; a lane marking segment is finally detected via piecewise curve fitting to the lane marking segment. The algorithm is verified on synthetic and real-word images. Results show that the proposed method not only has real-time performance and a high accuracy in detecting lanes of varying appearance, but also offers convenient adaptation of detection both to light-reflective road surface and further types of lighting disturbances.

### Funded by

• Figure 1

Flow chart of proposed algorithm

• Figure 2

(Color online) Detection of the starting point for a multi-lane scenario

• Figure 3

Elimination of outliers for denoising. (a) Outliers are marked by squares before elimination; (b) result after elimination

• Figure 4

(Color online) Results on synthetic images used for testing. (a) Original image degraded by salt-and-pepper noise; (b) centroid detection; (c) lane detection

• Figure 5

(Color online) Detection results on inclined roads in the national vehicle testing ground under different lighting. (a) and (b) Original images; (c) and (d) detection results

• Figure 6

(Color online) Detection results for lanes with different shapes and interference. (a) Dotted lane; protectłinebreak (b) tilted road with dotted line on one side, full line on the other, and with large area of stains; (c) tunnel at night; (d) expressway at night; (e)$\sim$(h) detection results

• Figure 7

(Color online) Detection results for lanes with different binarization methods. (a) Original image; protectłinebreak (b) binarization for Bernsen [14]; (c) binarization for iteration [13]; (d) binarization for Otsu [11,12]; (e)$\sim$(h) detection results

• Figure 8

(Color online) Detection results for different detection algorithms. (a), (d) and (g) Original images; protectłinebreak (b) detection result for Hough transformation; (e) and (h) detection results for LS; (c), (f) and (i) detection results for proposed method

• Table 1   Detection results for synthetic images with different types of noise
 Number of synthetic images Noise type Number of correct detection Correct detection rate (%) White Gaussian noise 96 91.43 105 Salt and pepper noise 99 94.28 Random noise 93 88.57
• Table 2   Comparison of lane detection results for different binarization algorithms
 Number of images Binarization method Average processing time for binarization (s/frame) Number of correct detection images Correct detection rate (%) Bernsen 26.112 102 92.73 110 Iteration 12.904 101 91.81 Otsu 0.046 105 95.54
• Table 3   Comparison of lane detection results for different detection algorithms
 Lane shape Number of images Detection method Average processing time (ms/frame) Number of correct detection images Correct detection rate (%) Straight 140 Hough transformation 906.30 109 77.86 Proposed method 269.64 136 97.14 Curve 150 LS 933.67 97 65.33 Proposed method 376.89 139 94.67

Citations

Altmetric