Comparative Analysis of Visual-SLAM Algorithms Applied to Fruit Environments

Argentinian fruit activity has been in crisis in the late years due to several factors, including the shortage of workers and high costs associated to the production. The development and incorporation of new tecnologies in order to automate and enhance the productive processes are urgent and unavoidable. In this line, this work is framed in the development of an autonomous navigation system based or artificial vision for a fruit platform for pruning and harvest labors. In particular, an experimental evaluation of Visual-SLAM strategies for stereo cameras applied to the estimation of the trayectory of a self-propelled platform in a typical fruit environment of the Patagonia region is presented. This proposal analyses the performance of 3 well-known publically available methods of the state-of-the-art, in order to conclude about their applicability in this particular environment. The results are based on image sequences captured in the installations of INTA Alto Valle (Instituto Nacional de Tecnología Agrícola Alto Valle) in real light and weather conditions, and they are contrasted with GPS-RTK measurements.