Semantics-Aware Next-best-view Planning for Efficient Search and Detection of Task-relevant Plant Parts

To automate harvesting and de-leafing of tomato plants using robots, it is important to search and detect the task-relevant plant parts. This is challenging due to high levels of occlusion in tomato plants. Active vision is a promising approach to viewpoint planning, which helps robots to deliberately plan camera viewpoints to overcome occlusion and improve perception accuracy. However, current active-vision algorithms cannot differentiate between relevant and irrelevant plant parts and spend time on perceiving irrelevant plant parts, making them inefficient for targeted perception. We propose a semantics-aware active-vision strategy that uses semantic information to identify the relevant plant parts and prioritise them during view planning. We evaluated our strategy on the task of searching and detecting the relevant plant parts using simulation and real-world experiments. In simulation, using 3D models of tomato plants with varying structural complexity, our semantics-aware strategy could search and detect 81.8% of all the relevant plant parts using nine viewpoints. It was significantly faster and detected more plant parts than predefined, random, and volumetric active-vision strategies. Our strategy was also robust to uncertainty in plant and plant-part position, plant complexity, and different viewpoint-sampling strategies. Further, in real-world experiments, our strategy could search and detect 82.7% of all the relevant plant parts using seven viewpoints, under real-world conditions with natural variation and occlusion, natural illumination, sensor noise, and uncertainty in camera poses. Our results clearly indicate the advantage of using semantics-aware active vision for targeted perception of plant parts and its applicability in real-world setups. We believe that it can significantly improve the speed and robustness of automated harvesting and de-leafing in tomato crop production.