Grasp Detection under Occlusions Using SIFT Features

Figure 8

Grasp detection results on household objects under occlusions: (a) detection results of a garage kit, the first row of (a) is the input information and template information, the target object on scene image is occluded by other objects, and the template image only contains the target object and its grasp configurations; the second row shows the SIFT feature pairs obtained by our matching algorithm, and we connect each matching SIFT features using a blue line; and the last row is the detecting results with multiple-rectangle above and most suitable rectangle below, and all the results avoid the occluded parts. (b), (c), and (d) have the same layout. We connect the matching features using blue lines.

(a)

(b)

(c)

(d)