Feature Extraction And Recognition Of Human Sitting Posture Based On Machine Vision

As an important branch of the computer field, machine vision technologyhas a great use of space in intelligent surveillance, human-computer interaction,sports analysis and virtual reality, etc. Therefore, machine vision technology hasimportant significance for study. In this paper, the posture of human body whichis based on machine vision technology were studied, combined with little car,which is used as simulated intelligent wheelchair.In order to detect different postures, two aspects were studied in this paper:detection of face and track of colored mark points. Because of the headmovement would be drove by the upper body while under sitting posture,therefore, the detection of facial movement is used to judge the movement ofupper body torso. However, the trend of leg movement could be judged bytracking the movement of leg mark points. The face detection method which isbased on AdaBoost algorithm was studied in this paper for face detection andtrack. This method can both satisfy the real-time of face detection and theaccuracy rating of detection is also relatively high. Among them, it includes Haarfeature, integral image, weak classifier, strong classifier and cascade classifier.Detection speed can be boost effectively by using cascade classifier. However，KLT(Kanade-Lucas-Tomasi) algorithm was selected in this paper, among theresearch of face track, it realizes real-time and accurate track of face target. Themethod of tracking mark points is used for the judgment of leg movement. Findthe center location of the color connected domain in the images, and then to trackthe movement of leg, where the mark points are, with the1010black rectanglemark. The influence of experimental results are worked out under different heightand different light condition by analysis of experimental data.The detection and track of face is applied as the control of intelligentwheelchair, the little car which is simulated as intelligent wheelchair was designed, including Dc motor driving module, obstacle avoidance module,photosensitive module and so on. To judge the facial movement by computer, themovement of little car, which is to simulate intelligent wheelchair, can becontrolled by command via serial port. The experimental result shows that themovement of simulator can be controlled quite stable by this method, therefore, itput forward a valid solution for the movement control of intelligent wheelchair, italso solve the problems like basic trip of disabled people(unable to use thehandle control of wheelchair).