The Biological Feature Detection And Recognition Of The Hand Image

With the rapid development of' intelligent robot technology and its wide application in the manufacturing system, the intelligent man-machine coordination assembly is desired to deal with the demand of high flexible in assembly system for electromechanical products. In the mechanical assembly of man-machine interactive system, biological structure feature detection is desired in process of hand assembly image and it makes the robot system have fully recognition of the hand assembly action and intention from operator, all of these have important theoretical significance and engineering application prospects for the improvement of intelligent robot. Under the deep analysis, extraction and recognition of the assembly biometric features in the assembly image, the hand shape recognition in the assembly is realized based on the structural biological feature infomation.On the basis of the random characteristics in the knuckles, the image preprocessing has made with homomorphic filtering and using Gaussian process model to study the characteristic of image category in the different location of hand joints, As input with the two kinds of model ·likelihood value marked, the finger joint target image has tested according to the results of estimation directlty, which implemented the hand joint recognition based on stochastic clustering characteristics.By measuring skin concent rati on, the image segmentation is used YCbCr color model,through grayscale averag,e variance instead of grayscale average in the Otsu algorithm which is used dynamic threshold to complete the skin color segmentation and make morphological operations of the rest of the burr and fill the void. On the basis of biological structure characteristics of hand, the color of skin information is associated with more biological characteristics of hand image, the finger fitting is used by b-spline, the position detection of fingertips, root and the palm is implemented using convex hull detection and curvature criterion,at the same time, combined with the test results of knuckles image to rich more information in the local characteristics of finger area, the biological characteristics of hand image is more effective detected and identified.Using the method of feature matching to calculate parallax image and improve the area matching algorithm, with corresponding to the region matching algorithm which is based on parallax, as the image parallax value in the left block for the current position, the right one is searched for information and direction until get the corresponding matching parallax figure.Combining the region matching algorithm with feature matching algorithm, with considering of skin color segmentation and the feature of hand biological, the matching of hand joints position is realized effectively.