查看文章

Nowadays, harvesting robots play a vital role in agriculture. In this paper we aims to develop a machine vision system to locate the on tree mango fruits, which can be implemented in robotic harvesting as well as yield counting. We developed two frameworks to detect and count the on-tree mangoes. The first framework uses the k-means clustering techniques to group the image into different clusters where image is grouped into different clusters and then the cluster with mango objects are processed further. In the second framework, the image is converted into binary image format and then the erosion method is applied to find out the mango object in the original image. In both frameworks the detected mango objects are counted based on the centroid value. Then the results of both frameworks are compared. The input images are taken from Google images.