Research On Blood Cell Image Feature Extraction And Classification Based On PPED

Blood analysis is an important item in medical testing, and identification of different types of cells provides an important basis for clinical diagnosis. The existing detection methods are mainly flow cytometry and image analysis technology, The use of flow cytometry equipment is bulky and expensive, the existing image analysis techniques use high magnification microscopy to collect cell images, neither of these can meet the requirements for miniaturization of the testing equipment. The on-chip lens microscopic imaging system is a new image acquisition system. It is small in size, low in cost and widely used. However, due to the absence of an optical amplification system, the collected microscopic image resolution is too low for cell recognition And classification.In order to solve this problem, the main content of this paper is to extract the feature of low resolution cell image, and then complete the classification of white blood cells according to the extracted features. The main contents include three parts:First, the whole image of a single white blood cells in the segmentation, cell staining, the white blood cells because of the nucleus of the fate will be colored. Taking into account this feature, this paper uses an image segmentation method based on HSV color space conversion and a specific threshold to segment white blood cell images. The main steps are color space conversion, histogram equalization, threshold selection, noise removal and boundary tracking,segmentation effect is good.Then, the feature extraction algorithm of white blood cell image is extracted, and a feature extraction algorithm PPED (projection principal-edge distribution) is used to simulate the feature points in the cell image. The convolution of four directions is used to check and extract the feature points in the cell image. The extracted feature points are compressed to convert the white cell image into a vector of 64-dimensions. After analyzing the scale invariance, gray invariance and computational time complexity of the algorithm, the size and weight of the convolution kernel in the feature extraction process are modified. The modified feature extraction algorithm takes into account the rate of feature extraction And precision.Finally, the feature extraction algorithm is used to extract the feature of the white blood cell image, and the classical classifier k is used to classify it and compare with the results of the blood analyzer. The time of feature extraction is reduced by 17%, and the correctness of the three types of cells is more than 90%.