Application Of Digital Imaging Principle Component Analysis Of Urine Formed

Urine sediment analysis is an important basis for the analysis of clinical pathology, it has become a significant project for discussion. But when we finish the urine sediment analysis by hand, due to its complex composition, and daily huge urine analysis demand, manual analysis has been difficult to meet the requirements. With the appearing of urine sediment analyzer based on image, the problem can be alleviated. The instrument’s core algorithm of urine sediment analyzer is image processing and pattern recognition, image processing technology is mainly to transfer the original image information into feature, and recognition technology is mainly used to classify, then finally achieved automatic analysis of urine sediment.In this paper, for that core technology, we presented a urine sediment analysis algorithm based on SVM algorithm. The algorithm firstly uses the Gaussian low-pass filtering to the original microscope as the image pre-processing, then gains the single cell image via the image morphology.During the process of single cell image segmentation, this paper presents a single-cell segmentation algorithm based on nonlocal active contour model. The segmentation algorithm firstly gains the matching patches from the image, then according to the energy function and constraints entries, using the variational method for single cell image segmentation, then gets the intact contours of single cells. After the completion of the single cell image segmentation, we extract the geometry features, gray features and texture features based on Gabor wavelets from the region obtained in segmentation of cells.Finally, we put the above-mentioned characteristics into the trained SVM to finish the identification and analysis of urine sediment. In this paper, the urine sediment mainly is: BACT bacteria, NSE non-squamous epithelium, white blood cells and red blood cells. Experimental results show that the proposed algorithm can better identify the physical components of the four categories, test data accurate rate reached 81.25 percent. Main work done follows:1 we put forward a single cell image orientation algorithm based on image morphology, it mainly used for the single cell division, and it cuts microscope image into a plurality of single cell images.2 we propose a single cell segmentation algorithm via the nonlocal active contour model, it mainly gains the intact area of cells from the single cell images.3 through the single-cell area of the image feature extraction, we present an urine sediment analysis algorithm based on SVM algorithm.