基于帧差能量图遗传算法的自遮挡步态识别

Self Occlusion Gait Recognition Based on Genetic Algorithm with Frame Difference Energy Image

传统的步态识别方法在处理自遮挡步态识别问题时,通常由于从视频序列中分割出来的轮廓有噪声而不能很好地进行特征提取。为了解决这个问题,提出了基于帧差能量图(Frame Difference Energy Image,FDEI)的遗传算法(Genetic Algorithm,GA),首先采用数学形态学图像处理方法填平轮廓的漏洞并消除噪声,然后借助于步态能量图计算出步态图像的帧差能量图,接着从轮廓图像序列中提取出步态特征,最后,利用遗传算法完成步态的识别。在中科院自动化所-B(CASIA-B)步态数据库上实验验证了所提方法的有效性,实验结果表明,与几种先进的步态方法相比,所提方法在处理自遮挡步态识别问题上取得了更好的识别效果。

Usually, due to the noise contour segmentation from video sequences, the traditional gait recognition method can not be good for feature ex-traction in dealing with self-occlusion gait recognition problem. To address this problem, genetic algorithm based on frame difference energy image is proposed. Firstly, mathematical morphological image processing methods are used for holes remedy and noise elimination, then frame difference energy image of the gait image is compntered by means of the gait energy image, gait features are extracted from the silhouette image sequences. Finally, genetic algorithm is employed to complete the identification of gait. The effectiveness of proposed method is verified by the experiments on gait database B of Chi-nese Academy of Sciences Institute of Automation （CASIA-B）. Experiment ：results show that proposed method has better recognition efficiency on self occlusion gait recognition comparing with several latest gait approaches.