Local patch-based methods of object detection and pose estimation are promising. However, to the best of the authors’ knowledge, traditional red-green-blue and depth (RGB-D) patches contain scene interference (foreground occlusion and background clutter) and have little rotation invariance. To solve these problems, a new edge patch is proposed and experimented with in this study. The edge patch is a local sampling RGB-D patch centered at the edge pixel of the depth image. According to the normal direction of the depth edge, the edge patch is sampled along a canonical orientation, making it rotation invariant. Through a process of depth detection, scene interference is eliminated from the edge patch, which improves the robustness. The framework of the edge patch-based method is described, and the method was evaluated on three public datasets. Compared with existing methods, the proposed method achieved a higher average F1-score (0.956) on the Tejani dataset and a better average detection rate (62%) on the Occlusion dataset, even in situations of serious scene interference. These results showed that the proposed method has higher detection accuracy and stronger robustness.

The detection and pose estimation of objects in human demonstrations remain challenging yet crucial tasks. The increasing availability of red-green-blue and depth sensors makes it possible to synthetize local features of color and three-dimensional (3D) geometry, which are useful for processing a wider range of objects. However, existing methods fail to combine the inherent advantages of these two features. Moreover, pose refinement methods based on whole point clouds are often affected by occlusion and background noise. In this paper, feature points of the speeded-up robust feature and the fast point feature histogram were transformed into the same 3D space. After matching them separately, multimodal feature points were jointly used to estimate a coarse pose. Subsequently, the coarse pose was refined by aligning point clouds composed of feature points' neighboring patches. During the iterative closest point process, we selected corresponding points in matched local patches. In our first and second comparative experiments, F1 scores were respectively increased by 0.1349 and 0.1633, which verified the validity of our method. Finally, the third qualitative experiment showed that the proposed method is applicable to manipulated-object detection and pose estimation.