Combining Scale-Space and Similarity-Based Aspect Graphs for Fast 3D Object Recognition

Abstract

This paper describes an approach for recognizing instances of a 3D object in a single camera image and for determining their 3D poses. A hierarchical model is generated solely based on the geometry information of a 3D CAD model of the object. The approach does not rely on texture or reflectance information of the object's surface, making it useful for a wide range of industrial and robotic applications, e.g., bin-picking. A hierarchical view-based approach that addresses typical problems of previous methods is applied: It handles true perspective, is robust to noise, occlusions, and clutter to an extent that is sufficient for many practical applications, and is invariant to contrast changes. For the generation of this hierarchical model, a new model image generation technique by which scale-space effects can be taken into account is presented. The necessary object views are derived using a similarity-based aspect graph. The high robustness of an exhaustive search is combined with an efficient hierarchical search. The 3D pose is refined by using a least-squares adjustment that minimizes geometric distances in the image, yielding a position accuracy of up to 0.12 percent with respect to the object distance, and an orientation accuracy of up to 0.35 degree in our tests. The recognition time is largely independent of the complexity of the object, but depends mainly on the range of poses within which the object may appear in front of the camera. For efficiency reasons, the approach allows the restriction of the pose range depending on the application. Typical runtimes are in the range of a few hundred ms.

BOLD Features to Detect Texture-less Objects

“…Given the aforementioned limitations of descriptorbased object detectors, state-of-the-art proposals tackle the texture-less object detection problem by means of edgebased template matching [11,12,24,25]. One major merit of edge-based template matching is the ability to detect seamlessly both textured as well as texture-less objects.…”

“…As demonstrated in [9], though, the method can be harmed by partial occlusions. Another recent relevant template matching approach for texture-less object detection is proposed in [25], which however, unlike BOLD, requires full-3D object models to carry out the training stage.…”

Fast 6D pose estimation for texture-less objects from a single RGB image

“…Template matching approaches [15], [16], [9], [17], [18], [13] have been successfully used with texture-less objects: these methods create a set of template images of the object from different viewpoints, and estimate the pose parameters by finding the template that is more similar to the observed image of the object. Different similarity measures have been proposed such as Chamfer distance [15], [16], image gradient orientations [19], point-to-point edge correspondences [18], object appearance [17], or by using custom descriptor-based distances [8], [9], [10], [13], [11].…”

“…Different similarity measures have been proposed such as Chamfer distance [15], [16], image gradient orientations [19], point-to-point edge correspondences [18], object appearance [17], or by using custom descriptor-based distances [8], [9], [10], [13], [11]. Note that the accuracy of template matching approaches heavily depends on the size of the template dataset: the larger it is, the closer to the actual pose are the estimated parameters.…”

“…Although RAPID provides a fast and accurate pose estimation, it requires an initialization close enough to the actual optimum to be robust and avoid getting trapped in local minima. Notice that this is actually the case when a coarse prior pose is estimated by a detector [18], [17], or given by the pose at a previous frame of a sequence during tracking [20].…”