A Novel Visualization Tool for Evaluating the Accuracy of 3D Sensing and Reconstruction Algorithms for Automatic Dormant Pruning Applications

Fangda Li, Somrita Chattopadhyay, Shayan A. Akbar, Noha M. Elfiky, Avinash Kak; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2016, pp. 72-80

Pruning is one of the most essential procedures in specialty crop production. During winter, skilled workers have to remove certain primary branches of the dormant trees in order to increase the productivity of those trees. The process is particularly challenging as it needs huge workforce and costs, besides, the unavailability of skilled seasonal workers. In this paper, we introduce a novel multifunctional graphical user interface which operates in two different modes, namely, manual and automatic, and serve the following purposes - (1) validating 3D reconstruction algorithms, (2) automating the decision making process for selecting candidate branches to be pruned, (3) hence, providing an easy interface with any robotic pruners, (4) training unskilled seasonal pruners, and (5) disseminating the knowledge. The paper describes the operating modes of the proposed tool and highlights various features and functionalities of the software by conducting experiments on an exemplar apple tree.