局部视角无畸变的运动目标检测加速算法

A Motion Detection Accelerated Algorithm Based on Distortion Correcting of Local Field Angle

针对大视场视频监控系统球形视场畸变严重以及要求实时处理等问题,提出了一种球形视场内局部视角无畸变快速展映方法,实现了大视场条件下运动目标检测的加速。打破现有算法先完成图像畸变校正,后进行目标检测的惯例;仅对球形视场内运动目标所在的局部视角图像进行畸变校正和识别等操作,大大减少了计算数据量,降低图像处理的时间开销,在保证识别准确的前提下,满足了监控系统实时在线处理的要求。最后,对不同分辨率、不同视场角的摄像机进行了多组畸变校正及运动目标检测实验,并将实验结果与现有算法进行了比较。实验结果验证了所提算法对大视场图像采集设备进行无畸变运动目标检测的可行性和高效性(加速5倍以上),为低成本大视场视频监控系统的实时、准确的目标检测、识别奠定了基础。

In large field video surveillance system, distortion is serious, and real-time processing is required, a local field angle distortion correcting method in spherical field of view （FOV） was pvoposed, in order to accelerate the motion detection under the conditions of the large field of view. Breaks down the established rule, the processing, such as distortion correction, recognition etc. , is just on the moving objects in local field angle, which greatly reduces the data quantity and processing time. Under the premise of ensuring the accurate recognition, it meets the requirements of real-time online processing. Finally, experimentize the distortion correction and motion detection algorithms for cameras with different resolutions and different FOV, then compares the test results with the general algorithm. Experimental results demonstrate the feasibility and the efficiency of this algorithm （speedup 5 times）, so it laid the foundation for real-time, accurate target detection and identification of low-cost wide-angle video surveillance systems.