Combining color and depth data for edge detection

Edge detection plays an important role in many advanced robot applications. Sometimes the boundaries of objects should be extracted. This paper proposes an edge detection algorithm that combining both color and depth data. Firstly, the Canny operator is applied to the color and depth image to extract initial edges. Then the color-edges are optimized using depth information to eliminate non-boundary edges. The depth-edges are also optimized using color information to remove the edges caused by missing regions. At last, the optimized color-edges and depth-edges are fused together forming the final edge extraction. Comparative experiments demonstrate that, even when two objects have the same color, the proposed algorithm can extract object boundaries successfully while the non-boundary edges are eliminated in the same time.