Recognition Study on Positioning Line Transection of Nerve Slice Image

It is a very important premise of nerve slice images registration to recognize the positioning line transection in the image. On the basis of the feature analyzing of positioning line transection, this paper designs the training and testing module sets. Then the least square support vector machine (LS-SVM) is trained by the training module set and optimized by genetic algorithm (GA) according to the testing error. And the optimized LS-SVM is employed to recognize the positioning line transections whose center pixel´s coordinates are picked out as the control points for images registration. The result of experiments for recognizing the positioning line transections in 200 nerve slice images show that the recognizing accuracy reaches 91.75%, which is better than the RBF neural network and unoptimized LS-SVM´s recognizing results.