Classification of cultivated rice Seed MR219 and MR269 varieties based on colour features using machine vision technique

Abstract

Final Year Project (FYP) was carried out to extract the colour feature of cultivated rice seed MR219 and MR269 using image analysis. Mostly in rice seed factory, seed inspection need experienced worker to distinguish the quality of the rice seed and the method done was very time consuming. Hence, Machine vision system that consist of couple-charged device (CCD) camera was used in seed testing. The study starts by collecting quality rice seed from the nearby legitimate distributer. This study was focusing on the rice seed type of MR219 and MR269 because these rice seed was often used for rice cultivation throughout Malaysia. Next, the images of MR219 and MR269 was captured using CCD camera. In addition, the study was carried out to create a programming using LabVIEW 2013 to provide convenience to any users who related in the research field of colour feature extraction. Through experiments carried out, the colour diversity of rice seeds can be determined and classified depending on the parameters set. The colour features parameter that are extracted are red, green, blue, hue, saturation, value, and intensity. In the analysis of data, MATLAB Neural Network software was use for data research and classification. The neural network pattern recognition tool was used to create and train network, and evaluate the performance using mean square error and confusion matrices. Studies conducted for the retrain 40 hidden layer show the MSE value of 1.885E-01 and the accuracy is 76.7%.