摘要
提出一种带种子补偿的时空背景差分高速公路车辆检测算法.由于高速公路场景的特殊性,首先基于混合高斯模型的背景差分、相邻帧差法以及邻域背景差分法的结合消除光照变化、场景扰动对检测结果的影响;然后通过基于HSV颜色空间的阴影消除判断并消除被误检为车辆的阴影;最后通过跟踪种子补偿去除差分方法造成的空洞.实验验证,论文提出的方法能有效去除光照变化、阴影等环境因素的影响,提高车辆检测的准确度和识别率.
This paper proposes a method of particular scene of highway vehicles detection algorithm based on the space-time background difference with seed compensation. Firstly,considering the particularity of highway scene,we used GMM background difference,frame difference method and the neighborhood background subtraction to remove the scene disturbance and some of the effects of environmental factors. And then,we determined the position of the shadow through HSV color space and eliminate shadows. Finally,we used the tracking seed compensation to remove the voids. By these experiments,we proved the highway vehicle detection method can effectively eliminate the influence of illumination change,shadows and other environmental factors,and thus improve the accuracy of vehicle detection and recognition rate.
出处
《四川师范大学学报（自然科学版）》
CAS
北大核心
2016年第3期456-460,共5页
Journal of Sichuan Normal University（Natural Science）
基金
四川省科技厅科技支撑项目(2012GZX0090和2014GZX0002)
