US 12,462,609 B1
Systems and methods for privacy-aware weapon anomaly detection via integrated object recognition and skeletal motion analysis
Liwei Ding, Miami, FL (US); and Mohammadhadi Amini, Miami, FL (US)
Assigned to The Florida International University Board of Trustees, Miami, FL (US)
Filed by Liwei Ding, Miami, FL (US); and Mohammadhadi Amini, Miami, FL (US)
Filed on Jul. 16, 2025, as Appl. No. 19/271,242.
Int. Cl. G06V 40/20 (2022.01); G06V 10/26 (2022.01); G06V 10/764 (2022.01); G06V 20/52 (2022.01); G06V 40/10 (2022.01)
CPC G06V 40/20 (2022.01) [G06V 10/26 (2022.01); G06V 10/764 (2022.01); G06V 20/52 (2022.01); G06V 40/10 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A system for privacy-aware weapon anomaly detection via integrated object recognition and skeletal motion analysis, the system comprising:
a processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
a) detecting human figures and weapon objects in real time within each frame of an input video using an object detection model;
b) classifying detected individuals as armed or unarmed by applying a proximity-based target separation technique to associate detected weapon objects with corresponding human figures, and generating a weapon confidence score indicative of the likelihood that a given individual is armed;
c) anonymizing unarmed individuals by executing a head segmentation module configured to perform a facial masking operation on human figures classified as unarmed based on an absence of associated weapon objects;
d) identifying abnormal motion patterns by applying a skeleton-based diffusion model to motion data extracted from detected human figures, and generating an anomaly score indicative of a deviation from normal behavioral patterns;
e) applying a feature refinement module to the weapon confidence score and the anomaly score to generate a refined weapon confidence score and a refined anomaly score, the feature refinement module comprising a plurality of filtering operations; and
f) applying a fusion process to the refined weapon confidence score and the refined anomaly score to generate a final anomaly score and classifying the input video as normal or abnormal by comparing the final anomaly score to a predetermined threshold.