融合运动和统计特征的静态目标检测

Combining Motion and Statistical Features for Static Object Detection

为了解决大部分时间处于相对静止状态目标的智能监控,提出一种融合运动和统计特征的静态目标检测方法.该方法采用行列错位减图像的帧差来提取目标运动特征,根据目标模型和候选区域的统计特征匹配检测目标,利用运动特征和模板的相似性度量动态更新模板.通过积分图优化特征提取及对强光抑制,提高了算法的实时性和鲁棒性.

To detect the static object which is relatively static in most time, a method for static object detection was proposed by combining motion and statistical features. The image of inter-frame difference, which was formed by one image subtracted from the image with a pixel offset in row and column, was used to obtain the motion feature. The statistic feature of the target region and candidate region was used to detect the target. The template of the target was updated according to the motion feature and similarity between the target and candidate region. By optimizing extraction of the statistic feature with the integral image and elimination disturbance of strong light, the performance of real-time and robust was improved.