Classification of Defects in Injected Parts Through: Moments of Image and Multilayer Perceptron.

Quality control at industry is an important factor to find the desired specifications of the manufactured products, this is often achieved manually with the support of the work staff. It implies that some defects are not detected by workers due to fatigue. This article describes a technological proposal, applied to the quality control of injected parts, using vision machine to give support to quality control. In this article we achieve obtain a set of PVC tee in a previously were environment it for four different faces and two types of parts found: defective and non-defective. Then image processing where we obtain the characteristic vector and using feed forward artificial neuronal network based on a multilayer perceptron mentioned images are classified account to the type expected. An n training procedure and is performance to classifi images are discussed in detail in the contect of quality control.