Detection and Counting of Lemons using Artificial Vision and Tracking Techniques for Real Time Harvest Estimation

Abstract

Nowadays, estimating the amount of fruit harvested is an important process for a farmer, providing a significant tool for making decisions about production. This work aims to automate the counting of lemons in real time during the harvesting process, using low processing vision equipment and low resolution cameras mounted on a lemon harvester. Different techniques were used, such as color-based image processing, vegetation and contrast index, mathematical morphology and object tracking based on the Kuhn-Munkres algorithm. To evaluate the performance of the algorithm, six harvest videos with a resolution of 640x480 pixels were tested, resulting in a success rate of over 95% between the visual count and the count provided by the algorithm.