Statistical Video Tracking of Pomegranate Fruits

Abstract

Efficient location of fruits in the trees is the most important criterion of an automatic robotic harvesting system. The main challenges faced in the development of the robotic harvesting arm are accurate identification of the fruits in dense foliages and detection of the occluded fruits. This paper proposes a statistical technique that accurately detects and tracks pomegranates in trees. A k-means clustering methodology is implemented as a preprocessor step followed by image analysis algorithms that finally locate the fruit accurately under varying conditions of illumination, human intervention, camera view etc. The simplicity of the technique is attributed to its real-time capability to detect and track fruits in a video of fruit-bearing trees. The ground truth data were computed on 25 video sequences and the system was tested on another 25 videos. The detection efficiency is as high as 96.3%.