Identification of impurity in wheat mass based on video processing using artificial neural network and PSO algorithm

摘要

The presence of weed seeds and other impurities in the wheat grains affect the identificationquality of the wheat grains. This study explores the possibility of identifyingthe wheat in the wheat grain mass and estimates the amount of impurities inthe wheat mass based on the video processing combined with the artificial neuralnetwork (ANN) integrated by particle swarm optimization (PSO) algorithm. After preprocessingthe video of the mass movement on the conveyor belt, 35 shapes, color,and textural features were extracted from each grain sample in the image in the presenceof the MATLAB software and image processing (IP) toolbox. The data obtainedfrom the IP section were categorized into two approaches. The first one, purposesthe identification of the wheat grains in the grain mass, and the other one, identifieseach components in the wheat mass. Both of them employs the hybrid ANN-PSOalgorithm to achieve the highest classification accuracy and the lowest error value.According to the results, ANN with the architecture of 36-10-10-5 and ANN withthe architecture of 36-8-8-5 overtakes the other architectures with the highest accuracyand performance values for training (100 and 100%) and testing phases (100and 86.66%), respectively for the first and second approaches. Accordingly, basedon the results of the hybrid ANN-PSO algorithm, the highest classification accuracywas 98.62% and 97.77% (for training and testing phases, respectively) and 76.08%and 73.1% (for training and testing phases, respectively) related to the first and thesecond approaches, respectively. Finally, the video processing using ANN can be consideredas a powerful approach for identifying the impurities in the wheat grain mass.Practical applicationsOne of the major problems in wheat and similar mass products is the presence ofimpurities inside the total mass, which drastically reduces the quality and the marketabilityof the product. Conventional methods of detecting the amount of impurities inthe grain mass are always accompanied by difficulties such as lack of precision, timeconsuming, and tedious. The use of new technologies such as machine vision and artificialintelligence have been considered as the important tools in this regard. One ofthe practical applications of this study is to detect the amount of impurities in wheatgrain mass at the inlet of silos and in pricing on grain masses. The system proposed inthis study can accurately detect and report impurities in wheat grain mass.