Research on garbage classification and recognition method based on improved GMM model

Classification and recognition is the precondition of garbage classification. The traditional manual classification and recognition is inefficient, and it is difficult to meet the requirements of garbage classification in modern life. At present, there are still many problems in the research of intelligent garbage classification and recognition, such as fuzzy classification of solid waste, low recognition of moving objects, slow running of program and so on A hybrid model (GMM) for garbage classification and recognition is proposed, Four types of solid waste graphic database including hazardous waste, recyclable materials, kitchen waste and other waste were constructed, and the data set was annotated. Aiming at the situation of huge amount of garbage data, complex shape and even partial overlap, this paper proposes to study the recognition problem of garbage classification on the slow-moving pipeline. By comparing GMM, SVM and k-NN models, GMM model is selected and improved. GMM model will consume large amounts of system resources when building background model. At the same time, it is easy to produce the phenomenon of ghosting when the long-time stationary target turns to motion. In order to solve the above problems, the number of Gaussian distribution function established by the pixel is simplified, which can be adjusted automatically. The moving target detection algorithm is improved and the ghosting is eliminated. Using Halcon software to build deep learning system to classify and identify the target, using multi thread reading to improve the efficiency of image processing and program execution.