Efficient, parallel, and invariant shape recognition algorithms for complex industrial tasks

The task of recognizing complex objects in an industrial environment is considered. Two massively parallel algorithms for shape recognition that are based on the notion of horizontal and vertical distances between two points on the XY plane are introduced. It has been shown theoretically that these algorithms provide close to perfect invariant geometric characterization of every possible image. The algorithms are applied to the task of distinguishing between commonly cited industrial parts to demonstrate their efficacy in real-life scenarios. Issues related to rotation by arbitrary angles and magnification in the discrete domain are addressed in detail. The proposed schemes provide theoretically predictable, invariant, and efficient shape recognition and discrimination