Improving the process of detecting visual defects in chestnuts for export purposes

Abstract

An important economic activity in countries such as Peru, Bolivia, and Brazil is the trade of nuts, such as chestnuts. Before export, quality control is required, including the assessment of ripeness and the detection and identification of defects, damage, or diseases. This process is based on the product's external characteristics, using color, shape, size, and texture descriptors. The aim is to automate the detection and identification of visual defects in objects such as chestnuts. The product is divided into two regions (dark and light) due to the color similarity between the object and the defects. Detection is performed through texture analysis in each region, using the Detect Defect algorithm (Alg. 2 and 3) and the First-Order descriptor (Alg. 5). To identify specific defects, color, size, and texture descriptors are applied through Color and Size Segmentation (Alg. 4). The proposal was implemented and tested on a database, achieving an efficiency rate of 97.90% and a processing time of 17 to 25 ms per image, outperforming the algorithm of the project (Proy.PIPEA_134, 2013), which reached an efficiency of 91.06% and 43 ms of processing.