Research Of Urinary Sediment Image Acquisition And Analysis System

As one of the three routine inspection, urine examination in clinical physicians determine disease has important reference value。 There are a lot methods of urinary sediment examination, but the different instruments and methods are likely to cause a few results present false negative or false positive,which leading to difficult diagnosis of clinical doctors.There are three methods in Clinical urine test:dry chemistry method, streaming urinary sediment automatic analyzer, urine sediment microscopy detection, etc.. Microscopic examination of the urine sediment is divided into automatic and artificial microscopy.This article have achieved a microscopic examination of the urine sediment system, which including the urinary sediment image acquisition, processing and the storage of image data and patients information,also flow processing part of the automatic control system part, finally added report printing function.Processing image after image collecting, first of all,extract the edge information of the image’s visible parts in order to make every visible part into separate clear domain, and then make the connected domain characteristic value input to BP neural network to carry on the visible part of the classification and identification.In terms of edge detection, first of all to convert color image into gray level in order to remove the redundant information of the image.Then use the median filtering method to remove salt and pepper noise.After using neighborhood filter, makes the image gray level transformation abated to get rid of the image acquisition count pool in the projection image.Then use Canny operator and Sobel operator respectively on image edge detection after the superposition of two images on average, in the processing of the image formed elements can form a complete connected domain. Finally fill the holes and corrosion expansion after the operation, the connected domain is filled to the white areas.Before feature extraction,the connected domains of the image should be extracted in turn first, then extract8characteristic value just as the area, perimeter, rectangle, round degrees, aspect ratio, seven moment invariants of Hu..Use the8characteristic value as input vector into BP neural network, the type of the output value is visible part. In order to improve the system operation speed, all the processing of the image uses OpenCV.The whole system is set up in VC++.net environment based on the MFC, using MFC technology to realize the workstation system.SQL Server is used to store patient information and image data. Interaction ODBC technology is the interface between VC++environment and SQL Server. Printing is still based on the MFC in packaging printing function. Finally, in the interaction with flow detection system is, serial communication technology is used. For the greatest extent without losing a serial port information and does not affect other operations, multi-threading technology is used.The system operations simple, fast, and is low cost.