计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (24): 227-233.DOI: 10.3778/j.issn.1002-8331.2008-0249

• 图形图像处理 • 上一篇    下一篇

融合时域和分水岭信息的车辆检测算法

马庆禄,唐小垚   

  1. 重庆交通大学 交通运输学院,重庆 400074
  • 出版日期:2021-12-15 发布日期:2021-12-13

Vehicle Detection Algorithm Combining Time Domain and Watershed Information

MA Qinglu, TANG Xiaoyao   

  1. School of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China
  • Online:2021-12-15 Published:2021-12-13

摘要:

在实时视频图像车辆目标检测时,为了克服行进中车辆背景噪声和阴影带来的准确率低、漏检率高等问题,提出一种时空融合和内外标记的分水岭车辆检测算法。通过相邻视频三帧差法得到的时域运动变化信息结合Canny算子得到的边缘图像相结合,得到时域掩模图像。利用文中提出的基于二次重构、内外区域标记、梯度修正的分水岭空域算法对运动区域及其周围区域进行分割,解决了一般分水岭算法的过分割现象。将得到的结果进行投影,以提高运动状态下车辆的检测精度。实验结果表明,在车辆背景噪声和阴影的影响下,该算法的检测效果仍然较好,车辆漏检率降低到4.90%,算法的准确性、鲁棒性和适应性较好。

关键词: 智能交通, 车辆检测, Canny边缘检测, 内外区域标记, 分水岭变换, 融合时空

Abstract:

In order to overcome the low accuracy and high omission rate caused by background noise and shadow of moving vehicles in real-time video image vehicle target detection, a watershed vehicle detection algorithm based on space-time fusion and internal and external marking is proposed. The time domain mask image is obtained in combination with time-domain motion change information obtained by the adjacent video three-frame difference method with the edge image obtained by Canny operator. The moving region and its surrounding region are segmented by using the watershed spatial algorithm based on quadratic reconstruction, inner and outer region marking and gradient correction, which solves the over segmentation phenomenon of general watershed algorithm. The results are projected to improve the detection accuracy of the vehicle in motion. Experimental results show that, the proposed algorithm detecting effect is still better, which reduces vehicle missing rate to 4.90%, and the algorithm accuracy, robustness and adaptability are better under the influence of background noise and shadow of the vehicle.

Key words: intelligent transportation, vehicle detection, Canny edge detection, internal and external area marking, watershed transformation, fusion of space and time