基于序列轮廓特征的步态识别研究

发布时间：2019-01-05 16:40

【摘要】：随着计算机视觉技术的发展,步态识别技术逐渐成为身份识别领域的一大研究热点。步态识别是根据人们走路的风格来识别个体身份的一种生物识别技术。与其它生物识别技术相比,它具有非接触、远距离、难以伪装、受外部环境影响小等优势,使它成为一种不可代替的生物识别技术。因此,本文深入研究步态识别技术。首先,研究步态图像的预处理技术。从视频中提取步态图像时,由于视频的背景比较简单,故采用背景减除法提取运动目标,进而将目标图像二值化。针对有噪声的图像,采用形态学的相关操作进行处理。由于目标图像的尺寸不统一,将图像归一化,得到归一化步态轮廓图。其次,针对单一轮廓特征识别率低的问题,提出基于轮廓特征的步态识别方法。该方法将步态区域面积、关键距离和质心到下半部分最外轮廓距离特征进行融合,识别效果显著。最后,为了提高步态识别率,体现步态的动态变化过程,提出基于序列轮廓变化特征的步态识别方法。提取序列步态关键距离、序列步态变化量和序列最大宽度特征。分析单一及融合序列轮廓变化特征的识别率,并与单一轮廓特征的识别率进行比较。最终将步态区域面积和序列步态关键距离特征进行融合,即基于序列轮廓变化特征的步态识别方法。该方法在穿大衣和携带背包行走状态下,取得不错的识别效果。
[Abstract]:With the development of computer vision technology, gait recognition technology has gradually become a research hotspot in the field of identity recognition. Gait recognition is a biometric technique to identify individuals according to their walking style. Compared with other biometric technologies, it has the advantages of non-contact, long distance, hard to camouflage and little influence from external environment, which makes it an irreplaceable biometric technology. Therefore, the gait recognition technology is deeply studied in this paper. Firstly, the preprocessing technology of gait image is studied. When gait image is extracted from video, because the background of video is relatively simple, the background subtraction method is used to extract moving object, and then the target image is binarized. For noisy images, morphological operations are used to process them. Because the size of the target image is not uniform, the image is normalized and the normalized gait contour is obtained. Secondly, aiming at the low recognition rate of single contour feature, a gait recognition method based on contour feature is proposed. The gait area, the critical distance and the outermost contour distance from the center of mass to the lower part are fused and the recognition effect is remarkable. Finally, in order to improve the gait recognition rate and reflect the dynamic process of gait, a gait recognition method based on the feature of sequence contour variation is proposed. The key distance of sequence gait, the variation of sequence gait and the feature of maximum width of sequence are extracted. The recognition rate of the variation features of single and fused contour is analyzed and compared with that of single contour. Finally, the gait area and the key distance feature of the sequence gait are fused, that is, the gait recognition method based on the variation feature of the sequence contour. This method has good recognition effect under the condition of wearing overcoat and carrying backpack.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】