We have developed a new method to estimate a Next Viewpoint (NV) which is effective for pose estimation of simple-shaped products for product display robots in retail stores. The accuracy of recent pose estimation methods using Neural Networks (NN) based on an RGBD camera decreases when few texture and shape features are obtained at a current viewpoint. Moreover, it is difficult for previous mathematical model-based methods to estimate effective NV because the simple-shaped objects have few shape features. Therefore, we focus on the relationship between object pose and NV. The more accurate pose estimation is, the more accurate NV estimation is. In order to utilize the relationship, we develop a new pose estimation NN that estimates NV simultaneously. Experimental results showed that the pose estimation success rate of our method was 77.3%, which was 7.4%pt higher than that of mathematical model-based NV calculation. Moreover, we verified that the robot using our method displayed 82.5% of products.
Mizuno et al. (Thu,) studied this question.