Design of a Size Sorting Machine Based on Machine Vision for Mexican Exportation Mangoes

In this paper, we fully design the hardware and software for a system that sorts Mexican mangoes for exportation. We use a conveyor belt, a webcam and a presence sensor along with the AC motor controller and the pc serial communication board. An image is taken by the webcam whenever a mango is detected passing through the presence sensor. Then, the acquired image is converted to grayscale and thresholding is applied to segment the mango. A morphological opening operation is applied to remove white noise produced mainly because of the mango shadow. Finally, all the pixels in the mango surface are counted and the result is compared with our database registers in order to classify the mango into one of ten Mexican standard categories. Experimental results show that the average sorting accuracy rate of the system is 89.6%, taking 3.75 seconds to classify each mango.