Internal Defect Detection in Mango X-ray Images using Image Processing and Machine Learning Techniques

The mango seed weevil is found in the world's major mango-producing regions and causes significant economic losses in mango fruit production. Many nations have identified the mango seed weevil as a dangerous quarantine pest, and its presence in exported mangoes can lead to the rejection of entire cargoes to stop the pest from spreading into mango-producing nations. It can also cause decreased seed germination, fruit pulp damage, and early fruit drop in mangoes. To prevent this menace, non-destructive evaluation methods need to be developed to detect the presence of seed weevils without cutting the mango fruit. A non-destructive method makes it possible to identify infested mangoes before they reach the market or processing facilities. This early detection helps prevent the spread of infestation, reducing overall fruit loss. For this experiment, 55 healthy mangoes and 5 defective mango samples were used in this work. A pipeline has been developed using image processing and machine learning-based techniques to process the X-ray scans of mangoes, automatically detect the presence of seed weevils inside the fruit, and display the final result on the screen. It also generates an imprint of the inner seed profile, depicting the presence of the infection inside it. In identifying the defective and normal samples, a 94% accuracy rate was achieved. At the end of the scan of mango samples, the generated report includes internal defect details regarding the seed weevil defect presence and its location in the mango sample.