Pharmaceutical Large Infusion Foreign Matter Detection Method Based On Clustering Joint Sparse Representation

Pharmaceutical infusion is the most common large-volume injection drug in the pharmaceutical industry in china,its production safety is closely related to the health of every patient.Therefore,in this thesis,a common infusion product is taken as an example to study the detection and identification methods of visible foreign matter,and the experiment proves that this method effectively improves the efficiency and safety of large transfusion product detection.Firstly,this thesis briefly describes the research background and significance of large infusion foreign matter detection,analyzes the difficulties in the detection of large transfusion foreign bodies using machine vision,and introduces the research status of the application of machine vision to large infusion foreign matter detection equipment at home and abroad company and the current research status of the existing large infusion foreign matter detection algorithms.Secondly,the design of the large infusion foreign matter detection system and overall procedure of the large infusion foreign matter detection was briefly described according to the analysis of target and interference.Thirdly,This thesis focuses on the detecting and clustering algorithm for large infusion moving targets.After analyzing the detection objects in this thesis,the overall design of the detecting and clustering algorithm for large infusion moving targets is given,In order to realize the detection of large infusion moving targets and eliminate the interference of stationary background noise on the detection of large infusion moving targets,the improved block principal component tracking algorithm will be used to detect the moving target in the sequence image.In order to obtain the motion trajectory of the moving target in the sequence image,the k-means clustering algorithm based block sparse matrix is used to cluster the moving target in the sequence image.And then the motion trajectory of the moving target is used to construct its motion trajectory space feature vector as the input feature vector of the classifier.The reliability and validity of the proposed detecting and clustering algorithm are verified through the analysis of relevant experimental results.Subsequently,Aiming at the shortcomings of the traditional sparse representation classifier algorithm in detecting foreign matter with high leakage and false detection rate,an improved hierarchical joint sparse representation classifier algorithm is proposed.In this algorithm,the motion trajectory space feature vector of the moving target is used as the input feature vector to identify the detected moving target object to eliminate the random noise interference.The experimental results of the two algorithms that are compared with each other verify that the hierarchical joint sparse representation classifier algorithm can reduce the leakage and false detection rate of the detecting equipment of large infusion foreign matter.Finally,In order to further improve the detecting speed of large infusion foreign matter detection algorithm,Based on the weighted residual eigenvectors proposed in this thesis and the traditional feature weighted support vector machine algorithm,the weighted residual support vector machine algorithm is proposed for the detection of foreign matter in large infusions.Experimental comparison result that the hierarchical joint sparse representation classifier algorithms is compared with the weighted residual support vector machine algorithm demonstrates that the weighted residual support vector machine algorithm effectively improves the detecting speed of foreign matter detection.