Research On Walking Posture Classification Algorithm In Foot Tracking

In the traditional footpath study,the footpath tracing method was used to deduce the walking posture of the criminal suspects.Most of the classifications of walking postures are based on the experience and professional knowledge of the civilian police when describing the suspect's walking posture,and there is no uniform specification for the description language.On the one hand,this kind of phenomenonwhich causes chaos and it is not conducive to the classification of walking postures.On the other hand,there is also no relevant research for the development of walking posture classification standards in public security practice and teaching.The Hikvision video capture device is used to collect normal human walking status.MATLAB is used to frame the collected video data,and a walking posture database is established,in order to explore the human walking posture classification criteria under natural walking conditions.Three methods are mainly used to explore the standard of walking posture in people's natural walking state.First,an angle analysis method based on the human body model is used.Photoshop is used to calculate the angle of the human head,trunk and arms,from the human walking silhouette and the frontal image in the database.Using the obtained angle data as a classification feature of walking postures,the K-means clustering algorithm is used to classify the angle data of the head,trunk and two arms,thereby the classification criteria for walking posture is obtained.Second,angle analysis method based on the human body part of the centroid connection is used.Background subtraction processing on the human walking silhouette image is used to perform morphological processing on the human body image.Dividing the binarized human body image into six parts of the head: neck,torso,legs,and hands,in order to calculate the coordinates of the mass center of each part.The angle between the center of mass of the head and neck,neck and torso,neck and hands is caculated according to the centroid coordinates.Finally,the K-means clustering algorithm is used to classify the obtained slope data,so as to obtain a walking posture classification standard.Third,the method of convolutional neural networks is used.The image of the human walking posture database is grayed out: The data is divided into training set and test set,then the category number is marked.Finally,the matgvnet toolbox in matlab is used to train the data in the vgg model and googlenet model.Then the convolutional neural networkmethod is used to test the classification result of the test set to classify the walking posture.After analysis,it can be concluded that the classification method based on the human body structure and the analysis method based on the centroid connection angle can classify the classification standard of the walking posture,while the method of using a convolutional neural network can not obtain a good classification effect.