A Combinatorial RGB and Depth Images CNN-based Model for Oil Palm Fruit Bunch Detection and Heatmap Localisation for a Visual SLAM System

The harvesting job of cutting and collecting fruit bunches in oil palm plantations remains the most labour-intensive processing cycle. introduction an autonomous vehicle to assist workers promises better productivity. Such a driverless requires software module known as simultaneous localisation mapping (SLAM) guide navigate autonomously. This work proposes visual SLAM system with distinctive capability detecting localising loose fresh (FFB) on ground using intelligent image processing. is equipped depth camera capable capturing RGB images concurrently. Two VGG16-based convolutional neural network (CNN) models are trained acquired dataset FFBs ground. output from combinatorial FFB detection model then fed into called RTAB-Map. By combining system, can plan for navigation safely, perform bunch pick-up tasks, avoid collision experiment results show that proposed CNN detect localise significant accuracy various lighting conditions.