Method And Application Of Human Posture Recognition Based On Kinect

Human posture recognition technology is one of the most concerned research topics in the field of artificial intelligence and computer vision.With the research and development of Kinect series depth sensing equipment,its unique depth imaging technology and skeletal joint point location technology have strong adaptability to the changes of illumination,texture and color.Therefore,the human body based on Kinect depth sensing camera has strong adaptability to the changes of illumination,texture and color.Posture recognition has attracted more and more attention and has broad application prospects in many fields.The main tasks of this paper are as follows:1.Analyse the hardware and software performance of Kinect device and its core technology,including depth imaging and bone tracking.Kinect 2.0 is used to collect and build a human posture database,which includes RGB,depth and human skeletal joint points.It consists of 8 kinds of common human posture images,totaling 3751.2.Analyzing the methods of human posture recognition based on machine learning,including hidden Markov model,decision tree,support vector machine and deep learning algorithm,comparing their principles and characteristics,and choosing the convolution neural network model in deep learning to recognize human posture image,then emphatically expounding the relevant theory of convolution neural network,laying a foundation for the later experimental work.Basics.3.Based on the human posture database collected in this paper,a convolutional neural network model is constructed and tested.The average recognition rate of the model in eight posture categories of three groups of data is 87.50%.On the basis of analyzing the shortcomings of the experiment,the data enhancement method is used to expand the training set of RGB data which produces over-fitting,which alleviates the over-fitting phenomenon to a certain extent.Under the same training set and test set,the average recognition rate of 95.56%can be achieved by using the migration learning algorithm based on Inception-v3 depth model.The experimental results show that the model has a high recognition accuracy.4.Based on Inception-v3 experiment of migration learning algorithm based on depth model,the application of fitness assistant system is realized through GUI design.