Detection of Floating Impurities in Filled Beverage Bottles Using Digital Image Processing Techniques

• Main Author: See, Alvin Arthur A.
• Format: text
• Language: English
• Published: Animo Repository 2015
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/4754
• Institution: De La Salle University

Some small and medium enterprise here in the Philippines uses manual inspection to check for contaminants inside filled bottled water. In the recent years demand for bottled water increases because many Filipinos believe bottled water is a healthier choice over carbonated drinks. Another factor that increases the demand for bottled water is due to the lack of access to potable water by many Filipinos in different areas in the country especially when a calamity stuck a certain region of the country. The proponent shows a proof of concept whether a low cost web camera can substitute an industrial grade camera for an inspection machine and if it can be used to make a locally available inspection machine that small and medium enterprise can afford. The inspection machine is composed of a web camera, a desktop computer, a proximity sensor, LED lights, and a Gizduino microcontroller while the program was created with the use of EMGU OpenCV and Visual C# .Net. An image capture will be triggered if the proximity sensor detects a presence of bottled water and the microcontroller will communicate with the desktop computer to command the web camera to take a snap shot. The captured image will be duplicated to be used for the detection of the moth wing and the detection of the moth body. The software will clean and process the captured images and analyze if there are any presence of contaminants or not. There were two experiments done, the first one is when the bottled water is stationary and after processing 30 images, results shows 76.67% reliable to detect moth wings and 90% reliable to detect a moth body. The second experiment was done with the use of a conveyor to simulate a production line and after processing 30 images, results shows 26.67% reliable to detect moth wings and 43.33% to detect moth body. On the GUI of the program, the captured image, processed image, and blob analysis are shown. The detected contaminants are enclosed inside a green box to indicate it is contaminated.