基于角点的监控摄像头干扰检测

提出一种新的实时检测摄像头干扰的方法,建立一种基于最小核值相似区SUSAN(Smallest Univalve Segment AssimilatingNucleus)角点的图像特征函数,由于干扰发生时相对的几帧图像与正常图像相比会发生显著变化,因此通过比较图像的特征函数值来检测干扰的发生。利用一种逻辑判断方法做到对突变和渐变干扰的全面检测,并且避免对无害行为的误检,使之更具鲁棒性。对比已有的方法,分别对不同场景及不同类型的摄像头遮挡和转动干扰进行了多组测试,实验结果显示,该方法明显优于其它算法,能够很好地检测出对摄像头的遮挡及转动干扰,并且很好地避免了对无害行为的误检。

TAMPERING DETECTION FOR SURVEILLANCE CAMERA BASED ON CORNER

In the paper the author presented a new technique for real-time detection of camera tampering and built a characteristic function of the image using SUSAN corner's information.Camera tampering will cause the correlated frames of video to be significantly different to the normal ones,so the camera tampering can be detected by comparing images' characteristic functions.By the use of logic judgement algorithm the entire detection of the abrupt and the gradual tampering is made,while it is much robust and insen...