A Preprocessing Method for Gait Recognition
The results of image preprocessing may directly affect gait feature extraction in the area of gait recognition. Due to the influence of light, shelter and other external factors of gait image, some problems such as loss of information, image shadows, and
- PDF / 1,237,298 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 24 Downloads / 219 Views
Abstract. The results of image preprocessing may directly affect gait feature extraction in the area of gait recognition. Due to the influence of light, shelter and other external factors of gait image, some problems such as loss of information, image shadows, and improper threshold of image preprocessing may occur. In order to solve these problems, an image preprocessing method of gait recognition is proposed. Firstly, background image is extracted by background modeling, secondly, the target profile is extracted by the direct difference method; thirdly, the shadow elimination based on the HSV color model is carried out on the target profile map; Finally, the complete target profile is obtained by threshold segmentation. Experimental results on CASIA_A database demonstrate that this proposed method is quite effective on both target profile extraction and proportion comparison with the real area. Keywords: Gait recognition
Image pre-processing Shadow elimination
1 Introduction With the development of computer science and technology, information security issues have become increasingly prominent, the identification of the identity of the technical requirements are getting higher and higher. Gait recognition has attracted more and more researchers because of its advantages of non contact, non aggression and long distance recognition [1, 2]. Gait identification is not considering clothing, camera angle, and background under people walk for identification. It is mainly divided into three steps: gaits detection, feature extraction and classification. Thus, gait detection plays a key role in gait recognition, and will directly affect the subsequent steps, and also determine the In gait image detection, there may be dynamic background changes, dark background light, low contrast of background and objectives (even the human eyes is difficult to distinguish), the shadows are shown by the sun or by the light, all these factors are making the target result not ideal, the real outline of the objective cannot be reflected even through the repair operation accuracy of recognition. Wang Liang has proposed LMEDS (Least Median of Squares LMEDS) method for background modeling in gait preprocessing, and indirect difference determination
Supported by Program for Liaoning Excellent Talents in University. © Springer Science+Business Media Singapore 2016 W. Che et al. (Eds.): ICYCSEE 2016, Part I, CCIS 623, pp. 77–86, 2016. DOI: 10.1007/978-981-10-2053-7_8
78
H. Shao et al.
threshold is used for binarization processing, morphology and the registration method based on contour edge correlation is used to further track the foreground region [3]. Although this method can get full target profile, but relevant registration method based on contour edge algorithm is much more complicated. Reference [4] uses a simple threshold segmentation and morphological processing to extract object contour map, but the distortion of contour is more serious. Reference [5] obtains the background model by subtraction method and gets the target image by the maxi
Data Loading...
