Research On Algorithm Of Segmentation And Recognition Of Urinary Sediment Images

The visible components in the urinary sediment, such as red blood cells and white blood cells, plays an important role in diagnosis and differentiation of kidney diseases, urinary tract diseases and infectious diseases in medicine. Therefore, the research has important significance for analysis and recognition of the urine visible components. Compared with regular examination methods, the automatic recognition system of urinary sediment visible components has high efficiency and accuracy. The visible components in the urinary sediment represent various and complex characteristics, and urinary sediment image still exists defocusing greatly, low contrast, uneven illumination, image blur and other issues, resulting in the visible components of urinary sediment images is difficult to obtain and recognize rapidly and accurately. In this thesis, urinary sediment image enhancement, segmentation, feature extraction and classification are studied, and an effective segmentation and recognition algorithms for the visible components of urinary sediment are demonstrated based on digital image processing and pattern recognition techniques.Firstly, the experiment and analysis for traditional image enhancement methods were carried out for urinary sediment image enhancement. The image combination enhancement method is proposed based on the advantages of each method. The experiment shows that the proposed method is more conducive to follow image segmentation from the efficiency and accuracy.Secondly, the experiment and analysis for traditional image segmentation methods were carried out for urinary sediment image segmentation. Aiming at the existing issues and difficulties of the urinary sediment image segmentation, the combination segmentation algorithm is adopted, and the adaptive segmentation algorithm of urinary sediment image is proposed based on watershed algorithm and Canny operator. First, the coarse segmentation of the image was carried out base on the Canny operator combined with morphological method. The urinary sediment image are classified and processed based on the area threshold and the complexity of the image. After image enhancement in combination enhancement method, the mask matrix calculated by using Canny edge detection is used as an accurate marker. Ultimately, the result of image segmentation are obtained with watershed segmentation. Experiments show that proposed algorithm of urinary sediment image segmentation has a superior performance and solve the problems existed in the traditional segmentation algorithm.Thirdly, analyzing and summarizing related theories and literatures were carried out for urinary sediment image feature extraction, and 27 feature parameters based on shape and texture are selected to describe urinary sediment. The foundation and prerequisites for urinary sediment images automatic recognition is laid.Finally, the VPMCD is theoretically proved that it can be applied in the urinary sediment visible components identification based on the analysis for the principle and process of VPMCD for the image recognition. The further improvement of the method is proposed: B-VPMCD algorithm based on optimization of VPMCD model by Bagging is proposed to enhance the stability and accuracy of the method. Based on the feature extraction,the urinary sediment classification algorithm based on B-VPMCD and PCA is proposed. Experiment from the fact show that the method proposed in this paper is effective in urinary sediment visible components recognition and meet the real-time and accuracy of the whole system, and is superior to the traditional VPMCD, BP neural network and SVM.The urinary sediment visible components automatic identification classification algorithm is applied effectively in automatic urinary sediment analyzer. The proposed system will gain great value in engineering application.