Detection Of Foreign Particles In Medical Solution Based On Image Processing

Abstract:Foreign particles mixed in medical solution will cause serious damage to patients and even endanger their lives. Compared to manual inspection, the automated foreign particles detection method based on machine vision has the advantages of high efficiency, high accuracy and low cost. The main results and contribution can be concluded as follows:1. A simple and effective method is proposed to demarcate the region of detection (ROD) according to the characteristics of acquired images. Then, take advantage of the spatial and temporal characteristics of foreign particles and background pixels, the spatial variance and temporal variance are used to control the filtering range of edge-preserving bilateral filter in order to estimate the background image.2. The Tsallis entropy thresholding method is extended to three-dimensional space to segment the foreign particles. Not only the current pixel gray value is taken into account, but also the mean value of its spatial neighborhood as well as the mean value of its temporal neighborhood, which can suppress noises effectively. In order to reduce the computation of three-dimensional Tsallis entropy, the differential evolution algorithm is used to optimize it.3. The geodesic distance, which can better reflect the intrinsic structure of the datasets, is utilized to take the place of Euclidean distance to measure the similarity of data points in our proposed motion continuity affinity propagation (MCAP). In addition, the motion inertia model is applied to modify geodesic distance. According to the moving continuity, we derive a variation of standard affinity propagation algorithm that includes a temporal constraint to enforce the temporal continuity of data points. The max-product algorithm is used to derive new iterative equations. The experiments show that the clustering result of motion continuity affinity propagation algorithm significantly outperforms standard affinity propagation. It can achieve better results both in the stability and accuracy of foreign particles detection and meet the requirements of production line.The achievements of this research can provide theoretical foundation for future application of automated foreign particles detection system.