| The Lingwu dates is a main planting cultivar and currently dominant varieties in Ningxia, the continuous development of Lingwu dates lead the development of related industries which use Lingwu dates as raw materials. Early grading way of Lingwu dates mainly rely on manual, with the increasing yield of Lingwu dates, the continuous improvement of labor cost and the development of machine vision technology, the research apply the machine vision technology in the field of fruit maturity classification.First, combine the related national standards, local standards, industry standards, enterprise standards and farmers'standard of Lingwu dates to formulate the maturity grading standards.Then, with the analysis of single channel and multi-channel of color features, the (R-G)/(R+G), (R-G)/ (G-B) and H mean components are finally determined as the maturity feature of Lingwu dates. Color feature, BP neural network and SVM are used to study the grading of Lingwu dates. And the recognition rate of color feature is 83%, classification accuracy rate of BP neural network is as high as 89.8%, but its parameter setting is complex, and the process takes a long time, support vector machine model is also used and its acccuracy rate is 87.5%, it is mainly used for small sample and it is faster than BP neural network.The algorithm based on theory of compressed sensing is also used to realize the maturity grading. The dictionary matrix is established, and then with the help of DALM algorithm to solve the problem of L1 norm, and the test sample is represented sparsely by the matrix to realize the maturity grading of Lingwu dates, the recognition rate is up to 94.5%.Finally, four algorithms are compared respectively from error recognition rate, classification accuracy and algorithmic complexity; the results show that compressed sensing algorithm is better than others. The software is also designed to realize the maturity grading of Lingwu dates on color feature, BP neural network, SVM and compressed sensing. |
PDF Full Text Request