Recognition of partially occluded two-dimensional objects

Recognition of partially occluded objects is essential for many industrial applications of machine vision. A fast and reliable algorithm based on subtemplate matching of boundary images is proposed to tackle the problem. Dynamic programming (DP) is employed in the algorithm to ensure an optimal consistent solution. With the weightings of the subtemplate determined off-line, the computational effort is roughly proportional to MN, where M is the length of image boundary and N is the number of subtemplates to be matched. Experimental results show that the algorithm will work for a wide variety of objects.