Gait recognition using normalized shadows

摘要: Surveillance of public spaces is often conducted with the help cameras placed at elevated positions. Recently, drones high resolution have made it possible to perform overhead surveillance critical spaces. However, images obtained in these conditions may not contain enough body features allow conventional biometric recognition. This paper introduces a novel gait recognition system which uses shadows cast by users, when available. It includes two main contributions: (i) method for shadow segmentation, analyzes orientation silhouette contour identify feet position along time, order separate and silhouettes connected such positions; (ii) that normalizes segmented silhouettes, applying transformation derived from optimizing low rank textures texture image, compensate changes view orientation. The normalized can then undergo algorithm, this relies on computation energy combined linear discriminant analysis user proposed outperforms available state-of-the-art, being robust acquisition viewpoints.