A new method for estimating the pose of objects for material handling

The ability to accurately estimate the pose of an arbitrary object using a single gray-scale image of the object even when the object is obscured by dirt, snow, or some object is investigated. A robust method is presented called the Pi4 method, which estimates the object pose from a single gray-scale image of a striped pattern on the object. This Pi4 method produces pose estimates which are insensitive to partial losses of the target pattern. The image processing and resulting pose estimates of a highly visible target pattern with computer added destructive obscuration are described. The experimental results demonstrate that the Pi4 method does not need the entire line to calculate line position. The asymmetric shape of the Pi4 pattern stripes prevents the method from failing. In general, random obscuration will not result in complete loss of a line, and as long as the line is not completely lost, the line fitting provides a very accurate estimate of the line's location.<>