Computer Vision-Based Corn Seed Quality Detection with Total Variation Deblurring in Conveyor Systems

Abstract

This study addresses crucial practical challenges in the agricultural industry by developing a computer vision system to assess the quality of corn seeds on a conveyor system with varying speeds, which is the main focus of this research. One of the challenges in this process is the motion blur effect caused by conveyor movement, which leads to a decrease in image sharpness and detection accuracy. To overcome this issue, this study integrates the total variation deblurring method in the preprocessing stage. Test results show that the application of the Total Variation (TV) deblurring model provides a significant improvement in the performance of the detection model, especially for the YOLOv8 model. The precision value increased from 0.850 to 0.878, recall from 0.859 to 0.873, and mAP50 from 0.910 to 0.926. Testing at conveyor speeds of 0.30 m/s, 0.35 m/s, and 0.40 m/s showed that the combination of YOLOv8 + TV produced more stable results compared to YOLOv11 + TV. Therefore, the application of the Total Variation Deblurring method has proven effective in creating a more reliable and accurate automatic inspection system for corn seed quality, with direct application in agricultural seed processing using a conveyor system.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint