AI/ML Object Tracking
The Army has a need to track accurately explosive hazards remaining in post-incinerated ammunition scrap after it is identified by x-ray on a conveyer belt to ensure it is safely removed at the end of the belt by a robotic arm.
For this effort, our Data Scientists developed a real-time object tracking algorithm that receives locations of hazardous material within an x-ray image, reidentifies those detections in a separate video feed, assigns and maintains consistent bounding box IDs and object properties using image processing techniques, tracks the object frame-by-frame using path predictions and object similarity machine learning (ML) models, and communicates the location of the scrap at the end of the conveyor to a robotic arm for removal.
- ML model to play the role between the x-ray detection of explosive scrap and the physical removal process at the end of the conveyor
- Provides autonomous tracking via reidentification of hazardous scrap detection to create a hand-off from the x-ray detection to the object tracking model
- Tracks all objects in the field of view, with frame-by-frame confirmation that tracks are following the same object over time
- AI/ML algorithm
- Object tracking model
- End-to-end system with little to no human interaction ensures safety and decreases human error
- Automated ML-model training is easily adaptable to other domains
- Computer vision
- Explosives and hazardous material detection
- Continuous moving object from multiple imaging and video sources