Autonomous smart farming system using FLANN-based feature matcher with robotic arm

摘要

Farming1 is an excellent niche for automation and robotic innovations. The manual farming process deals with many repetitive and tedious tasks. It makes it perfect for applying automation through agricultural robot concepts, AGRIBOTS, which can lead to more effective and efficient farming. The AGRIBOTS can cope with various agrarian tasks such as seeding, watering, harvesting, and monitoring. One of the challenges in autonomous farming is identifying the accurate harvest time. This study addressed the challenge for carrot farms by developing an optimized Fast Library for Approximate Nearest Neighbors (FLANN) feature matching algorithm with a randomized KD-tree index and a Scale Invariant Feature Transform (SIFT) algorithm to classify carrots and weeds images. This paper proposes an autonomous, innovative farming approach for small to medium-sized enterprises (SME) with a robotic arm and a claw that can move in the x-axis and y-axis on a gantry crane to precisely automating the seeding, weeding, and harvesting processes. In addition, a web application was developed for farmers to monitor crops conditions such as humidity, temperature, and soil moisture in real-time. They can also observe the whole automated process of planting with the growing period of the corps on the calendar. Finally, a prototype was built and tested successfully to evaluate the overall performance of the proposed AGRIBOT. Over one hundred samples were used to train the classification model that reached the Percentage of Correct Classification (PCC) of 77% and the precision of 88%.