中文摘要：

针对视频图像中形状匹配的局限性，即当待检测物体出现平移、旋转变化时识别目标需要很长的计算时间，提出了一种基于轮廓特征的运动目标识别方法．首先获取能自动更新的背景图像，采用背景减法提取运动目标的轮廓，然后运用其轮廓的边界不变矩特征和形态学特征，构建一个轮廓特征向量的模型，再分析比较待测运动目标轮廓特征向量与每类标准样本之间的欧氏距离，实现对运动目标的识别分类．试验结果表明，该方法具有识别精度高、计算量小、实时性好的特点．

英文摘要：

In view of the limitation of shape matching in video sequence, i.e. when an object was moving or rotating, it took a long calculation time, a recognition method of moving objects based on contour features was proposed. First, the background image was captured and automatically updated. The contour of moving objects was abstracted by background subtraction. Second, a model of contour eigenvector was built based on the edge invariance moment and morphologic features. At last, the recognition and classification of objects were realized by analyzing and comparing the Euclidean distance between the contour eigenvector and each standard stylebook. Experimental results show that this method has high recognition accuracy, less calculation time and strong real time characteristics.