基于Grab Cut算法和傅里叶描述子的人物背面步态识别

Gait Recognition of Figure in the Back Based on Grab Cut and Fourier Descriptors

提出基于Grab Cut图象分割算法和傅里叶描述子的步态识别方法。通过HOG+SVM算法提取出视频中的行人区域,对该区域使用Grab Cut算法进行分割得到行人二值化后的图像,然后应用傅里叶描述子描述步态特征,最后在识别方面采用最近邻分类器进行识别。此算法在中国科学院自动化所的CASIA数据库上进行实验,取得较好的识别效果。

Presents a gait recognition algorithm based on Grab Cut and Fourier descriptors. Firstly, uses HOG ＋ SVM algorithm to extract the pedestrian area in the video, then utilizes Grab Cut to achieve pedestrian binary image and describing the gait feature by means of Fourier descriptors, finally uses the nearest neighbor classifier NN for classification. Applies the method described above to experiment on gait database of CASIA provided by institute of automation, Chinese academy of sciences and getting a good recognition performance.