A Computer Vision System for Pallets Verification in Quality Control

Flexible manufacturing systems have gained their place in industries thanks to the possibility of being adapted to a particular task. A robotized palletizing cell is an example of such a system, which is responsible for organizing and stacking loads to be transported. A key part in this process is the pallet and if it presents a problem, the whole cell may stop. Furthermore, it is possible that manipulator robots, transported loads, and any other part of the cell suffer damage. To avoid these problems, the objective of this work is to develop an artificial vision system capable of selecting wood pallets that are in good condition, which means free of defects such as splinters, cracks and broken or out of dimensions elements. The artificial vision system has been developed based on image processing and machine learning techniques and presented better results compared to the past studies. Image processing parameters of Canny and Hough Transform have been adjusted, followed by comparisons among Support-Vector Machines with different kernels and their parameters. An accuracy of 100% in the classification has been obtained based on the Support-Vector Machine with radial basis function kernel.