Reducing the Estimation Error of the Measure of Proximity Between Objects in Pattern Recognition

When recognizing similar or close objects, the recognition reliability becomes very low, since the measure of proximity between objects is close to the value of the resulting error in the measurement of features. The article proposes an algorithm for solving this problem. In this algorithm, the parameters of a standard object are subjected to a large number of measurements in the learning mode. The parameter of the investigated object is measured by such a number that does not affect the speed of the system. During recognition, each measured parameter is compared with all measured values of the standard parameter. Thus, the number of repeated measurements is rather artificially increased. Since the comparison process is performed on the computer programmatically, this does not affect the speed of the system. This algorithm has been simulated on a computer and positive results have been obtained. The processing of the results showed that the proposed algorithm can significantly increase the recognition reliability.