Research On Gait Recognition Algorithm Based On Feature Fusion

As an important means to ensure information security,biometric identification technology plays an important role in different occasions.As the only biometric technology suitable for long distances,gait recognition technology has broad application prospects in the field of intelligent video surveillance.In recent years,more and more researchers have paid more attention to more and more gait recognition.Algorithms are constantly being proposed and improved.In this paper,the gait recognition algorithm based on feature fusion is studied.The gait recognition algorithm is studied from the following aspects:(1)Gait video sequence image pretreatment: The target detection method is researched on the target characteristics of human body,and different background modeling methods are analyzed.The background model is established by Gaussian mixture method,and the gait video sequence is processed by background subtraction method.Separating the human target from the background environment and extracting the binary target foreground image;The gait cycle sequence is extracted according to the aspect ratio of the human body region in the image,and the noise is removed by morphological processing,and the gait image is normalized.(2)Extraction of gait features: Using the refinement algorithm to extract the skeleton from the gait map,locating the key points according to the contour and the skeleton,and extracting the skeleton angle features as the limb motion characteristics;The gait graph of one cycle is statistically averaged to obtain the gait energy graph and dimension reduction,reflecting the spatial and temporal distribution of the gait.Constructing Hu moments with translation,rotation,and scale invariance in discrete case,describing the shape of the gait map;Feature extraction is carried out in three aspects: skeleton feature,gait energy map and Humoment.(3)Classification and recognition of gait features: Comparing the classification and identification of three single-step features in the neighboring method and support vector machine respectively,determining the parameters of an optimal classifier and classifier,and obtaining the recognition rate of three single features under the optimal classifier.(4)Gait recognition of feature fusion: Three gait features are fused in the feature layer for gait recognition,and the gait recognition performance of the feature fusion algorithm in normal walking,wearing coat walking and backpack walking state is analyzed.