Research On The Method Of Human Body Gesture Recognition Based On Infrared Image

Intelligent video surveillance system is a hot topic in the field of image processing and pattern recognition,it is of great significance on military and civilian industries.And the intelligent monitoring based on infrared camera has the advantage of realizing all-weather work whether at night or in harsh environments.In this thesis,based on the infrared intelligent monitoring system as the background,target segmentation,feature extraction and target recognition are studied for its key technology that the infrared human body gesture recognition method.In the part of the infrared target segmentation algorithm,due to the traditional regional growth algorithm needs to manually select the initial seed point,a method of combining visual saliency with regional growth is proposed to improve it.A visual saliency model based on spectral residuals is used to determine the salient region of the target,by processing the salient map to obtain the center of gravity of the target area,and the regional growth segmentation is carried out with the center of gravity as the starting point to realize the automation of target segmentation.In the part of the feature extraction algorithm,aiming at the disadvantages of poor texture information and low distribution of gray information due to the imaging characteristics of infrared image,extracting the shape features of the infrared image with geometric features,Fourier descriptors,and the Hu features,and according to the principle of stability and distinguish ability to optimize the extracted features.In the part of the target recognition algorithm,this thesis respectively uses KNN algorithm and SVM algorithm to recognize the gesture of the self-created infrared human pose image library.And based on DTW distance an improved algorithm for FKNN is proposed,increases the accuracy of recognition.The experiments verify the effectiveness of the algorithm used in this thesis,it has great reference significance for the application of intelligent video surveillance system.