Nowadays, chestnuts selection process in Peru is done by hand, therefore some important problems occur. People who work in this area could make a lot of mistakes because their personal situation or environment variables influence them. Fatigue, feelings, light conditions or working comfort are examples. For this reason the production may become slow and/or imprecise and think about increasing production needs to hire more employees and spend more money. In that sense, it is proposed the automation of the chestnuts selection process for industrial scale, where real time computer vision techniques are applied in order to detect products defects, that are analyzed by some external characteristics: Oval shape to detect the chestnut product and the same descriptor with color and size descriptors to detect the chestnut defects and their types. In this way, this approach allows to improve and increase the quality of the chestnuts selection for its exportation, reducing errors in the process operations. Experimental results show that the performance achieved in each chestnuts selection is 86.78 %.