| The computer-aided automatic classification of urine sediment images is of great importance for the clinical diagnosis of many kinds of diseases. There are kinds of equipment for urine sediment images recognition. This project is based on the 'urine sediment images automatically recognition system of LX3000'. We use BP Neural Network in cells classifying and counting. We present a data pruning method in *10 images recognition, also we research some feature selection methods for classifying. In *40 images recognition, we put multiply classifiers fusion in use.As the image segmentation method isn't very perfect now, we have no standard to label the image areas after segmentation, we are not professional clinical operators and the data amount is large, so it has many difficult data in the samples. This paper presents a Bagging method for data pruning. We use the method in *10 images classifying, it reduce the training time and training errors, also it improves the nets' generalization ability.Multiple classifier fusion, or combination, is a modern technique in pattern recognition areas. Through pertinently combining different information from varies of simple classifiers, the classification accuracy can be fairly improved and the difficulty of designing a single, high-accuracy classifier could be avoided. In recent years, fusion methods of many kinds have been widely used in the identification of human face, hand-written characters, remote sense images, etc., but relatively rarely studied in the medical image area.We pay more attention to neural network integration, and using majority voting method in classifiers fusion. After analyzing the correlation between the nets, we put the least information entropy method in processing the nets and selecting more suitable ones, and also in improving the performance of Boosting method. The simulation results show the fusion model is better than single classifier and traditional fusion methods. |
PDF Full Text Request