AI/ML in Automated Scrap Inspection

The Army requires an automated means to detect and discriminate for energetics remaining in ammo scrap. ILW provides a highly-accurate and fast two-model approach to automate scrap classification expeditiously and correctly. Using Convolutional Neural Networks and comparative Edge Detection, ILW’s approach offers 2x assurance for MDAS ammunition, where ammunition considered safe requires MDAS classification by both models. By fully integrating trained models into the industrial conveyor system, ILW’s approach uses real time imaging as input, classifies those images, and outputs a decision on whether ammunition contains explosive material or not. The decision is coded for hardware messaging to trigger conveyor activity to route ammunition the correct direction. This approach will successfully perform as good as or better than 200% human inspection in a small fraction of the time. By the end of Phase I, ILW will have a performant classification pipeline ready for collaborative efforts necessary to integrate and test with hardware vendors during Phase II.