摘要
为提高夜间环境下车辆检测的精度,提出一种基于亮度累加直方图的车辆检测算法,利用汽车尾灯的高亮特征检测自车前方车辆。通过统计大量的尾灯亮度信息得到分割阈值,由该阈值确定最大类间方差法的初始阈值。在亮度累加直方图中采用改进的最大类间方差法确定最佳分割阈值,并使用该阈值分割图像提取尾灯目标。结合尾灯的形状、位置和颜色等特征进行尾灯筛选和配对,以检测到的尾灯对为目标实现夜间车辆的检测。实验结果表明,该算法能够准确地分割出尾灯目标,对夜间前方车辆的检测率较高、适应性较好。
In order to improve the vehicle detection accuracy in the nighttime environment, a vehicle detection algorithm based on brightness cumulative histogram is proposed, which detects from the front vehicles via the highlight feature of the taillights. The initial threshold of Otsu method is obtained from a number of taillights statistical information. The bright objects are extracted from images, based on the improved Otsu method in brightness cumulative histogram. The characteristics such as the shape, position and color of taillights are combined to select and pair them. The front vehicles can be detected by the paired taillights. Experimental results demonstrate the accuracy of the proposed approach on taillights segmentation, and demonstrate the effectiveness and robustness of the approach on vehicle detection at night.
出处
《计算机工程》
CAS
CSCD
2013年第6期239-243,共5页
Computer Engineering
基金
国家自然科学基金资助项目(51175290)
博士后科学基金资助项目(2012M510421)
关键词
夜间环境
车辆检测
亮度累加直方图
最大类间方差法
初始阈值
最佳分割阈值
nighttime environment
vehicle detection
brightness cumulative histogram
Otsu method
initial threshold
optimalsegmentation threshold