Unsupervised Drone Video Correlation for Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing event monitoring algorithms struggle to reliably correlate video data from drones operating in various planes, from ground to air, due to unique spatial characteristics and lack of human intervention in feature extraction and correspondence establishment.
Innovation Solution
The implementation of unsupervised deep learning techniques to extract features and establish correspondences between videos captured by multiple drones, allowing for event monitoring in an unconstrained manner without human supervision, and storing a model to analyze future videos for anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing event monitoring algorithms are used to correlate video data from drones operating in various planes, then the system can process multiple drone videos, but the reliability of event detection deteriorates due to unique spatial characteristics and lack of human intervention in feature extraction
Solution Approach 1:
The system employs unsupervised deep learning techniques that enable the algorithm to automatically extract features and establish correspondences between drone videos without human intervention. The system self-trains on unlabeled data, allowing it to adapt to various drone planes and spatial characteristics autonomously, thereby maintaining reliability while processing multiple video streams
Solution Approach 2:
The patent transforms the approach by changing from supervised to unsupervised learning parameters, allowing the system to handle diverse spatial characteristics of drones operating in different planes. This parameter change enables the algorithm to generalize across varying conditions without requiring manual tuning or human-labeled training data
2Extent of automation
If unsupervised deep learning techniques are applied to learn a model from unlabeled videos, then human intervention is eliminated and automation increases, but the complexity of the learning process and model training increases
Solution Approach 1:
The unsupervised deep learning system performs self-training on unlabeled video data, automatically learning to extract meaningful features and establish correspondences without human guidance. This self-service capability eliminates the need for manual annotation while the system autonomously manages the complexity of learning from diverse drone perspectives
Solution Approach 2:
The system performs preliminary unsupervised learning to establish a baseline model before actual event detection. This preliminary action pre-trains the model on unlabeled data, preparing it to handle the complexity of multi-drone spatial variations, thereby reducing the burden during operational phase
Data Source
AI summary
In one example, the present disclosure describes a device, computer-readable medium, and method for performing event monitoring in an unconstrained manner using a network of drones. For instance, in one example, a first video and a second video are obtained. The first video is captured by a first drone monitoring a first field of view of a scene, while the second video is captured by a second drone monitoring a second field of view of the scene. Both the first video and the second video are unlabeled. A deep learning technique is applied to the first video and the second video to learn a model of the scene. The model identifies a baseline for the scene, and the deep learning technique is unsupervised. The model is stored.