Knowledge Acquisition System on Structured Models of Target Object to Take Environmental Changes Into Account

Abstract

When we recognize target objects in the real world, we are required the deal with abstract information of target objects such as cars or persons. In pattern matching techniques, enormous quantity of data are dealt with for describing the target objects. We propose new knowledge acquisition system which can describe abstract information of the target objects. This system treats the object as the combination of common elemental objects, and obtains feature property of each object as the knowledge for the object. It can realize data compression and improvement of versatility. The performance of recognition with obtained knowledge is affected by environmental changes. Therefore, this system changes its thresholds of the sharpness according to the environmental condition.