Multi-Camera Video Analysis for Blind Spot Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods struggle to accurately detect abnormal behaviors, such as shoplifting, in areas outside the imaging range of monitoring cameras due to blind spots created by overlapping or moving camera angles, leading to difficulties in identifying and alerting for potential incidents.
Innovation Solution
An information processing system that analyzes video feeds from multiple cameras, identifies correlations between objects and persons using scene graphs and attention maps, and determines abnormal behaviors by tracking changes in relationships between objects and persons across different camera frames, issuing alerts when such behaviors are detected outside the camera's imaging range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image recognition methods are used to detect abnormal behaviors, then the detection process is simple and fast, but the detection accuracy is low in blind spots of monitoring cameras
Solution Approach 1:
The system segments the detection task into multiple components: visible area detection by individual cameras, blind spot identification through overlay analysis, and abnormal behavior detection specific to blind spot regions. This segmentation allows the system to focus computational resources on critical blind spot areas while maintaining overall system manageability.
Solution Approach 2:
The system transitions from two-dimensional image analysis to three-dimensional spatial reasoning by creating an overlay map that represents the three-dimensional coverage volume of multiple cameras. This dimensional enhancement enables the system to identify and analyze behaviors in previously undetectable blind spot regions.
2Area of stationary object
If multiple cameras are deployed to cover more areas, then the imaging coverage is improved, but the complexity of analyzing and correlating data from multiple cameras increases
Solution Approach 1:
The system merges data from multiple cameras by creating an overlay map that combines the imaging ranges of all cameras into a unified spatial representation. This merging process automatically correlates data across cameras, identifying blind spots as regions not covered by any single camera, thereby simplifying the analysis of multi-camera data without losing coverage information.
Solution Approach 2:
The overlay map serves as an intermediary data structure that mediates between raw camera feeds and behavior analysis algorithms. Instead of directly correlating data from multiple cameras, the system uses the overlay map to represent coverage areas and blind spots, which then guides the abnormal behavior detection process in a simplified manner.
3Reliability
If the system focuses only on visible areas in camera ranges, then the analysis is straightforward, but abnormal behaviors in blind spots cannot be detected
Solution Approach 1:
The system performs preliminary analysis to generate the overlay map and identify blind spot regions before conducting abnormal behavior detection. By pre-processing the spatial data to determine where blind spots exist, the system prepares targeted detection zones in advance, making the subsequent behavior analysis more focused and effective rather than attempting to analyze all areas uniformly.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
An information processing program causes a computer to execute a process including acquiring a video that is captured by one or more camera apparatuses identifying a relationship for identifying a correlation between an object and a person included in the video by analyzing the acquired video determining whether the person has performed an abnormal behavior on a product on an outside of an imaging range of the camera apparatus based on the identified relationship and giving an alert based on a determination result on whether the person has performed the abnormal behavior on the product on the outside of the imaging range.