Scientists are starting to use sonar to help computers recognize shapes in a similar way to how bats use echolocation.
Shape recognition is key in the development of computer vision, where a computer performs human-like activities, and any kind of software that requires image analysis. Computers recognize shapes by analyzing different aspects of an object and comparing them to stored items on a database. Computers must first simplify the object in a representation --- called a shape descriptor --- that contains the most relevant information for its identification. Shape recognition techniques are used widely in archeology, architecture, medical imaging and facial-recognition software programs.
Time Series Technique
The time series shape recognition technique does not rely on an object's size or rotation. It relies on the analysis of an object's contour through time. The computer takes several snapshots of the object through time and uses autoregressive model parameters to categorize and compare the object's contour with other objects in its database. Autoregressive analysis is a mathematical tool used in statistics and signal processing that uses sample data to predict future results.
Moment Invariants
Moment invariants are a popular method of describing the contour of a shape. Once a computer analyzes the moment invariants of a shape, it can compare it with objects with similar moment invariants in its database. Medical science, for instance, uses moment invariants to identify the similarities of protein binding sites.
Sonar Images
Scientists and engineers use high-resolution 3-D sonar to recognize the shape of objects. Sonar instruments collect acoustical information on an object in a similar way to how a camera collects information from the light that bounces off it. Once a shape matrix is developed from the data, it is processed using shape-recognition algorithms. This method allows the use of 2-D and 3-D shape-recognition algorithms.
Tags: moment invariants, object contour, recognize shapes, shape-recognition algorithms, through time