基于改进的YOLOV5算法对ADB汽车大灯的外界环境检测

Environment Detection of ADB Automobile Headlamps Based on Machine Vision and Deep Learning

摘要:

针对远光灯交汇会影响汽车驾驶员的视觉注意力，导致汽车驾驶员夜间行驶安全难以得到保障的问题，研究基于机器视觉及深度学习的ADB汽车大灯的外界环境检测方法。 通过机器视觉的CCD相机采集ADB汽车大灯外界环境图像数据，利用数据筛选方法剔除采集到的图像数据中干扰光源数据，依据路况特征差异，划定ADB汽车大灯外界环境检测目标区域后，通过深度学习算法检测外界环境目标车灯光源，结合扩展卡尔曼预测各目标车灯光源轨迹，当车辆前方有车灯光源经过时，ADB系统及时调整汽车远光灯对应区域灯珠亮度，减少在高速行驶时因远光灯交汇对汽车驾驶员的视觉影响，保障汽车安全行驶。实验结果表明，该方法可有效剔除各类干扰光源，准确检测目标车灯光源，且目标车灯光源轨迹预测结果与真实结果非常接近，可精准完成ADB汽车大灯的外界环境检测。

Abstract:

Aiming at the problem that the intersection of high beam headlights will affect the visual attention of automobile drivers and make it difficult to ensure the safety of automobile drivers driving at night, this paper studies the external environment detection method of ADB automobile headlights based on machine vision and deep learning. Acquire the image data of the external environment of ADB's automobile headlights through the CCD camera of machine vision, use the data filtering method to eliminate the interference light source data in the collected image data, delimit the detection target area of the external environment of ADB's automobile headlights according to the difference of road conditions, detect the light source of the external environment target through the depth learning algorithm, and predict the track of each target light source in combination with the extended Kalman. When there is a light source passing in front of the vehicle, The ADB system timely adjusts the brightness of the lamp beads in the corresponding area of the high beam light of the car, reducing the visual impact on the car driver due to the intersection of high beam lights when driving at high speed, and ensuring the safe driving of the car. The experimental results show that this method can effectively eliminate all kinds of interference light sources, accurately detect the target light source, and the trajectory prediction results of the target light source are very close to the real results, which can accurately complete the external environment detection of ADB automobile headlights.