Research And Implementation On Hand Segmentation In Complex Environment

Image segmentation is a difficult but very meaningful work. In the field ofimage processing and computer vision, it is difficult to find a unified imagesegmentation method, which makes image processing and computer scientists needto propose different image segmentation methods for different applications andproblems.With the development of computer automation technology, biometricidentification technology as a stable, secure, fast identification technology came intobeing, and continued the rapid development. Especially with high stability, highrecognition rate palmprint identification is the leader in this field. However, itsdevelopment constraints of certain palm collection environment.In order to break through such restrictions, complex environment basedpalmprint identification system comes into being. Hand segmentaion in complexenvironment is one of the most difficult problem in the entire palmprintidentification system with free environment and free attitude. This work is to discussthe hand segmentation in complex environment and to build hand segmentationsystem.Firstly, this work establishes a skin color model based on artificial neuralnetworks. Skin color model calculates pixel by pixel and generates a probabilitymap. Probability map is binarized using fast maximum entropy method to determinethe threshold. This part is the rough hand segmentation step in the wholeprocessings.Then, this paper presents a vote map based boundary extraction method. Themethod detects edges in each color channel using the Sobel operator and combinededge results based on the mechanism of neighbour voting.Finally, hand segmentation is completed by combination of rough segmentationand the boundary, a series of morphological operations and hole filling operation.The experiment was conducted on80640*480sized image library andobtained a sensitivity of96.52%, specificity of96.92%of the experimental results.In addition, this work provides some information on a hand segmentation systembased on the the algorithm. This system segments a hand in less than0.5S inreal-time mode.