Research On The Algorithm Of Urinary Sediment Images Automatic Recognition

Urinary sediment inspection is an important method of clinical diagnostics. The traditional microscopical inspection which is watched by one eye is easy to not only increase doctors' burden but also bring up artificial error. Furthermore, the image can't be transformed and processed by this means. It is very difficulty to remote-transmit, make precise measurement quickly and access conveniently the microscopical image. But management and analysis of urinary sediment based on computer system can increase efficiency of clinical inspection greatly and relieve the burden of doctors in clinical standardization of disease diagnosis and manage present provide & history information of clinic. The main task of this paper is the research on automatic recognition algorithm of urinary sediment images by digital image processing technique.Several major contents were presented in this paper. According to the traits of urinary sediment image such as more targets, complex background and so on, an algorithm based on canny operator is selected to segment sediment image more efficiently after analyzing and comparing all kinds of algorithms of image segmenting. Canny operator has the disadvantage during the edge detection of urine sediment image, which would not obtain the dithreshold adaptively and make the detected edges not closed. To improve this disadvantage, the candidate edge obtained after non-maxima suppression are dilated by using a small structure element for getting closed edges later, and a method of getting dithreshold is improved that the high threshold is determined according to the greatest gradient value of the image and the low threshold is determined according to the local gradient value, which is adaptive. Experimental results show that the improved Canny operator can automatically segment the urine sediment image with high accuracy and strong robustness to the noise. In order to separate the images of overlapping cells in the binary image which will infect the result of feature parameter picking and recognition, an algorithm based on chain code difference is presented in this paper. Experimental results show that the algorithm works well. To integrally describe sediment, five texture parameters, such as energy, inverse difference, entropy, inertia, absolute value, are extracted based on nine morphological and statistical parameters which can describe the feature of urinary sediment. All of this lay a foundation for urinary sediment images automatic recognition. At the end the BP neural network method has been adopted to classify urinary sediment, by which a high recognition rate was obtained.