Date of Publication :11th July 2023

Abstract: In many parts of the world, rice is the main source of food, and the quality of the grain affects both consumer health and the financial interests of agricultural traders who make offers to farmers based on the quality of the rice kernels they have harvested. When trading rice, the buyer must first assess the rice's quality before deciding on the purchase price. In the past, this employment relied heavily on manual labour. Professionals are needed to hand choose and weigh various varieties of rice from samples that were chosen at random. Calculating the weight ratio of faulty rice in the total sample size is how rice grades are determined. Rice quality estimate by hand is time-consuming since skilled inspectors must recognise, pick up, and carefully weigh each defective kernel individually. This study suggests a novel approach for automatic quality estimation of rice kernels in an effort to address these issues that the existing literature is unable to address and to replace human labour-intensive operations with automated procedures. The main categories of novel rice kernel faults being detected in this initiative include broken kernels (BR), spotted kernels (SP), yellow-colored kernels (YC), mass-chalky kernels (MC), and partial-chalky kernels (PC). This experiment used a multi-stage classification strategy to do multi-classification of rice flaws, allowing a single kernel with dual defects to be discovered and then classified using Yolo v8. This approach allowed for the detection and classification of rice kernels with diverse sorts of defects.

Reference :