A novel diagnostic scheme for the classification of pistons

摘要

In this paper, a computer vision system for diagnosing pistons during the process of manufacture is designed. In fact, it is almost always readily possible to fit the incorrect piston to an engine because a number of similar piston types exhibit very slight differences but with the same overall diameter. Within a family, the difference is very subtle and may be simply a change in the shape of the bowl in the crown of the piston. Hereby, a vision based measurement system is designed to confirm the identity of the piston just after it has been fitted to the engine assembly. Structured laser lines are employed to obtain depth information on the piston crowns. A unique calibration technique involving an optical system is introduced in the scheme, in order to achieve precise 3D measurements. Since some pistons have shining surfaces, the reflection on the crowns influences the detection of the laser line, hence the dispersion of the laser lines increases the difficulty of locating the laser lines. The concept of fuzzy sets is employed to describe the features of the pistons to enable the measurement system to be more tolerant to the various pistons and to be able to obtain more accuracy in the recognition scheme.