关键词: 车辆工程, 车道线识别, 图像处理, 阈值分割, 遗传算法, 车道偏离系统, 安全性

Abstract: Driverless vehicles need to perceive the environment through the vision system, and lane recognition and monitoring at front is an important part of driving route planning and driving safety monitoring. Aiming at the problem of real-time and adaptability of lane-line image recognition, a lane-line image threshold segmentation optimization algorithm based on the combination of maximum inter-class variance method and genetic algorithmwas proposed. The proposed algorithm determined the first generation population by binary coding, took the maximum inter-class variance calculation formula as the fitness function, and used the variance between classes as the fitness value. The threshold of image segmentation was calculated by genetic algorithm and the image processing was performed by MATLAB programming, which effectively eliminated noise, protected the details of the image and improved the accuracy, fitness and recognition speed of lane-line image recognition. Then, according to the lane-line feature, the lane-line detection and tracking algorithm was used for fitting. The lane departure algorithm was used to judge whether the car deviated from the lane line during the driving process. The proposed algorithm was transplanted to the vehicle image processing chip, and the correctness of the proposed algorithm was verified by real vehicle experiments.Finally,the real-time data of the current lane departure was obtained by the vehicle intelligent terminal and sent to the cloud server.Based on these data, the degree of lane departure was judged and the lane departure warning was realized.The experimental results show that the obtained lane departure data is in good agreement with the actual vehicle driving state at that time.

Key words: vehicle engineering, lane identification, image processing, threshold segmentation, genetic algorithm, lane departure system, safety