Evaluation of three spherical characterization algorithms.

*(English)*Zbl 0803.68119Summary: Object recognition algorithms can be tailored to recognize specific objects based on their geometric properties. This study demonstrates three algorithms that can quickly identify and characterize round objects such as blood cells, microscopic oil particles, or trapped air bubbles as seen by a video image processing system. The first algorithm, the most traditional of the three, the Full Perimeter Algorithm (FPA), uses a steepest intensity gradient walk about the perimeter of the object. The Single Chord Algorithm (CSA) is able to operate on, and characterize, a round object if only a portion of the object is in view but is operationally slowed because of floating point calculations. The Double Chord Algorithm (DCA) is optimized for integer arithmetic and yields the smallest computational time of the three but is the most easily mislead by dirt or other non-circular objects. The best solution is an intelligent blending of the three algorithms governed by a set of empirical rules that can determine dynamically which method should be employed for a given object.

##### MSC:

68T10 | Pattern recognition, speech recognition |

68U10 | Computing methodologies for image processing |