基于仿射不变矩的动物运动检测与跟踪

Animal Motion Detection based on Affine Invariant Moments

矩技术因为其具有数学上的简明性及多样性而得到了广泛的应运，如在计算机视觉和模式识别等方面。结合帧差法在运动目标检测中运算速度快的特点，提出了一种基于隔帧差分区域结合仿射不变矩的运动目标检测方法。该方法通过隔帧帧差方法先提取运动区域。然后计算待检测的运动目标的不变矩特征量，最后利用神经网络方法来对特征量进行识别。试验结果表明，该方法能较好地检测出运动目标，并且克服了对突然停止目标会出现检测检测不出来的弱点，同时提高了运算速度。

Due to its appealing simplicity in mathematics and versatility, the moment technique is widely used in many areas, such as computer vision and pattern recognition. Combining with the high operation speed of frame difference detection technology, the moving target detection algorithm, based on afline invariant moments and frame difference area, is proposed. Firstly, the algorithm picks up moving region by using the distance frame-difference method. Then it calculates the invariant moment characteristic quantities of the detected moving tar- gets. At last, identification to these quantities will be achieved by neural network. Test results show that the method can detect moving targets without missing, overcome the weakness that the target would be undetectable when suddenly stop, and at the same time, the method improves the computational speed.