AI emerges as a new weapon against the multi-billion-dollar illegal wildlife trade, offering a 92% accuracy rate in detecting trafficked marine species in airport baggage.
Research Method
A study published in Frontiers in Ocean Sustainability has demonstrated a method for using artificial intelligence to detect trafficked marine wildlife in X-ray baggage scans. The technology was tested on the detection of shark fins, seahorses, and sea cucumbers.
Researchers developed algorithms to identify the targeted species within 3D X-ray CT scans. To create a training dataset for the neural network, scientists collected 68 samples of deceased marine animals, including specimens from actual wildlife trafficking seizures. The samples were scanned using a 3D X-ray machine to build an image library.
The study utilized 298 individual scans. To simulate real-world smuggling conditions, some scans included samples concealed within items such as tin containers, clothing, and toys. Researchers applied a Threat Image Projection technique, where images of smuggled goods were digitally added to clean baggage scans to test the algorithm. The system is designed to flag suspicious items for subsequent human inspection.
Performance
In testing, the algorithm achieved an overall detection accuracy of 92% for the three species. Reported success rates for individual species were 95% for shark fins and 96% for seahorses, though one source reported a 95% rate for seahorses. The detection accuracy for sea cucumbers was recorded at 86%, with one source providing a figure of 85%.
The algorithm generated false positives at rates of 2% for shark fins and 1% for sea cucumbers. One source reported a false positive rate of 9% for seahorses, contributing to an overall false positive rate of 13%.
Context and Limitations
Illegal marine wildlife trafficking is estimated to be a multi-billion-dollar annual enterprise. The trade poses a threat to endangered species.
The study’s authors noted that the algorithm is not a replacement for existing detection methods, including human operators and biosecurity dogs. They also stated that many airports currently lack the 3D CT scanning equipment necessary to implement this technology. The system is intended to complement, not replace, these existing methods. The algorithms may occasionally miss items or produce false positives, necessitating verification by personnel.