Classification of Ripeness of Coffee Fruits Using Support Vector Machine Based on Computer Vision

This research implements computer vision to detect and classify the ripeness level of coffee fruits. The main contribution of this research is to improve the image quality that is degraded due to blur caused by the movement of objects on the conveyor belt. Image quality restoration is done using a Wiener filter on images detected using the YOLOv4 model. This allows the classification performed using a Support Vector Machine to achieve higher accuracy. Experiments were conducted using coffee fruit images taken using a conveyor moving at 60rpm and 90rpm, using a Logitech Brio camera with a resolution of $\mathbf{1920}\times \mathbf{1080}$ pixels. The total image samples used were 300, of which 80% were used as training data. The experimental results show that the blurred image repair using Wiener filter produces a PSNR value of 27.25dB for the ripe class, 27.2dB for the semi-ripe class and 27.57dB for the unripe class, where the classification of the ripeness level of coffee fruit without using an improved image with an accuracy of 93.3% compared to the classification of the ripeness level of coffee fruit using an restoration image reaches an accuracy of 98.3%.