Rice fissure detection using computer vision

Rice fissure detection using traditional manual method encounters with some problems such as great subjectivity, high random, low efficiency and bad repeatability. A rice fissure detection system was developed after researching the illumination characteristic of fissured rice in the platform of VC++6.0 software. Single rice images were picked up from originally obtained mass image with methods of binary treatment, region marking. Rice fissure characteristics was emphasized by gray transformation, then the curve of row-mean gray value of single rice kernel image was drawn and a filter algorithm was designed to smoothen the curve. Subsequently, an algorithm was designed to detect rice fissure based on row-mean gray value of single rice kernel image. Finally, a detecting experiment was carried out with six groups of specially-selected samples and five randomly-selected samples of five varieties of rice kernels such as Jinyou974, Feiyou600, Gangyou182, Zhongyou205, 89-94. The results show that the detecting accuracy of the system for specially-selected samples and randomly-selected samples were 98.37 per cent and 97.88 per cent, respectively. It could help to improve rice quality detection with computer vision theoretically and practically.