An appealing option for businesses looking to save costs on component inspection is automated visual inspection (AVI). Camera systems for component inspection allow for less manual work and maybe better output rates. For the automobile sector, where product lifespans are often short and where suppliers are under growing pressure to provide flawless components, this is of paramount importance. However, setting up an AVI system may be complicated, particularly for businesses that have little expertise with vision systems. This causes training periods to be lengthy and the advantages to be under-realized. The goal of the work presented in this thesis is to create and build an image-processing algorithm that may aid in shortening the required training period and adjusting for external factors, such as surface lighting, that can affect the accuracy of the results. The software is able to learn the inspection process from samples of excellent and poor components thanks to the use of "intelligent" algorithms (particularly neural networks and fuzzy logic). The user is not only giving the data necessary to generate an accurate classification, but also, indirectly, the predicted variance in the photographs via the provision of sample images. This implies that the intelligent algorithm does not need being informed the maximum permissible variance, in contrast to more conventional systems. The algorithm was put through its paces using data from a customer in the industrial sector as well as photographs created in the lab. Results indicate that training times may be drastically reduced by switching to an example-based method. Classification performance was found to be on par with that of conventional threshold-based methods for pictures with a clear pass/fail distinction. The neuro-fuzzy system somewhat outperforms other methods when dealing with small discrepancies. In order to implement and test in an industrial setting, a user interface was designed.