Research Based On External Characteristic Parameters Of Tomato Classification

With the development of machine vision technology, this research utilize this technology successfully extracting the RGB images of common kind of tomato surface defects, identifying them, making the judgment Intact tomatoes were graded in the size and the weight according to the national standard. The main contents and results of this research were as follows:(1) Experiment platform was established and improved for this study. The lighting system was perfected, CCD sensor was chosen properly, the background for taking pictures was selected.The areas of calyx,areola,decay,and mechanical injury etc were extracted successfully through the use of local recognition, image enhancement, image gray-scale adjustment with the combination of analysis of binary image and color image. This method was applicable in the common condition, and the experimental results showed that this method was good for common surface defects extraction.R,QB values and parameter values of extracted parts were chosen as the indicators to identify tomato surface defects. The experiment results showed that the recognition correct rate were high. The areola recognition accuracy reached96.1%, calyx recognition accuracy was100%, intact tomato recognition accuracy was100%, the decay fruit recognition accuracy was91.3%, the mechanical injury area recognition accuracy was82.3%.(2) The algorithm was built based on Matlab platform, using mathematical morphological operation and image local property operation to establish identification algorithm for the tomato. Linear model, second order polynomial model, power model were established through analyzing the relations between tomato external parameters and weight. Weight prediction model was established through regression analysis of the external characteristic information. The experiment results showed that the weight was highly related to its area, perimeter, maximum inscribed circle diameter and minimum circumscribed circle diameter. Test results showed that the multivariate linear prediction model was the best, its determination coefficient (R2) was0.9267, standard deviation (S.E) was4.32. its mean relative error was1.535%, and the mean absolute error was3.260g. The experiment results show that the program can quickly predict tomato weight through tomato external features.(3) The equivalent diameter d1was chosen to take the place of maximum transverse diameter as the indicator of grading. The experimental results showed that tomato size classification accuracy was98.2%.(4) A software system based on Matlab for freshwater fish variety identification and weight prediction was developed. The functions of the software system included image pretreatment, feature extraction, freshwater fish variety identification and weight prediction etc. Test results showed that the software system was operational simple and stable operation.