Mangifera indica real-time quality classifications using codebook segmentation and mass-size correlation equations

Abstract

Indonesia as a tropical country, has high rate production of mangoes and it's potentially rise up the country income. Unfortunately, the Indonesian mangoes export rate are poor. One of the reason is damage caused by mechanical post-production sortation. The digital sortation using codebook segmentation model and mass-size correlation equations are able to reduce the damage that caused by mechanical sortation. This research purposes are to do codebook segmentation model and mass-size correlation equations and digitally sort the mango by its mass approximation. The data that used in this research are the maximum length, width, and thickness of mango. Those measured values are required to approximate values of mango mass using the mass-size correlation equations. The approximate mass will be classified into three classes according to the Indonesian National Standard. This program able to classified the mangoes with 95.83% of average accuracy, but the MSE value of each classified mangoes (Class 3 and 4 based on SNI) are 1091.619344 and 61.75204226. Overall, the codebook algorithm is able to recognize the object and count the desirable measure even though the program still recognize shadow as an object.