Schistosoma Cercariae Image Recognition via Sparse Representation

The paper proposes a method of schistosoma cercariae image recognition via sparse representation(IRSR). In the method, all the schistosoma cercariae image training samples compose the dictionary for sparse representation. For each test sample, its projection coefficient in the dictionary is computed and the category which has minimal residual value is assigned to it. We also investigate the effect of the size of the training set on the recognition rate. At last, IRSR is compared to the k-nearest neighbour method(KNN), back propagation neural network method(BP) and support vector machine method(SVM). The experimental results show our method can achieve 97% recognition accuracy, which is the best recognition result in all the above methods´. IRSR provides a novel and effective scheme for schistosoma cercariae image recognition.