Automatic cacao pod detection under outdoor condition using computer vision

Abstract

This study aims to detect and count cacao pods on trees using a 4K resolution drone. The image was taken with the difference in distances and illumination in the plantation area which affect the color variation of the object's surface. The data variation is a challenge in this study. There are 85 images of cacao fruit on trees were taken with three variations of distance, which are 50 cm, 100 cm, and 150 cm. For each distance, all images processed through several stages, i.e. image quality improvement using image enhancement at the preprocessing stage, image segmentation using the K-means method, and BLOB analysis method to detect and count the number of cacao pods on the tree. The result shows that the best distance between the camera and the object is 50 cm with an average accuracy of 93.3%.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint