In order to stand out from the competition, the only alternative for automotive industries, including wheels manufacturing industries, is to grant lot of importance to their products quality. In recent years, most of the customer returns of defective products are due to appearance defects, which are associated with the wheel aesthetics. These defects are located on the outside of the wheel and are mainly related to the quality of the painting. In general, the defect detection process is a manual process conducted by operators, which is subjective and difficult due to the complicated wheels surface. This paper proposes to design a fully automatic computer vision system to inspect the whole surface of the wheel. It is proposed to use four cameras placed over the production line. A diffused lighting system is considered in order to illuminate homogeneously the whole surface of the wheel. Each wheel is designed with specific parameters that define its form and geometry. For defect detection, an original adaptive linear parametric model is proposed. This model is sufficiently general to describe any type of wheels. The adaptivity of the proposed model makes it sufficiently accurate to describe the inspected wheel surface while detecting small defects by using an optimal statistical hypothesis test. In addition, the computer vision system monitors in real time the wheel coating intensity. Using a parametric sequential method, this monitoring allows the detection of any abrupt changes, i.e. small decreasing amount of paint, that would reveal a sudden problem in the painting process.