Automatic Method for Measuring Object Size Using 3D Camera

This paper proposes a method to automatically measure object size with high accuracy by a structure-light based stereo camera system. The approach includes four steps namely preprocessing, object detection, key points extraction and depth interpolation before size calculation. The RGB and depth frames are firstly aligned in the preprocessing step. Then, the object is detected by depth threshold and key points are extracted by Shi-Tomasi corner detector combined with our proposed key point extraction algorithm. An interpolation algorithm is developed to handle incorrect depth at the edge of objects. Last but not least, the object dimensions are calculated by Euclidean distance through the 3D coordinate of the key points. Experimental results of three-object measurement show that average accuracy is about 99% which is close to those of manual and theoretical methods. These also outperform the results of recently published papers.