Differentiating Abandoned and Removed Objects via Edge Orientation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video analytics systems struggle to distinguish between abandoned object events and removed object events, as these events are indistinguishable using traditional background subtraction methods, leading to inefficient alerts and resource-intensive computation.
Innovation Solution
The Orientation Transition Comparison (OTC) method compares the transition of edge orientation for each boundary block in a video frame, classifying regions of change as abandoned or removed based on the change from more random to less random or less random to more random patterns, eliminating the need for threshold settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional background subtraction methods are used to detect regions of change, then both abandoned object events and removed object events can be detected, but the two event types cannot be distinguished from each other
Solution Approach 1:
The patent segments the boundary of the detected region into multiple boundary blocks, and further segments each boundary block into multiple super-pixels. This hierarchical segmentation allows analysis of edge characteristics at different granularities, enabling differentiation between abandoned and removed objects through statistical comparison of edge orientations across segments.
Solution Approach 2:
The patent transitions from analyzing single pixel values to analyzing edge orientation distributions across multiple boundary blocks. By introducing the dimension of edge orientation analysis and comparing statistical properties across spatial segments, the system gains the ability to distinguish event types that appear identical in traditional background subtraction.
2Measurement precision
If pixel-based boundary strength methods are used to differentiate event types, then some differentiation is achieved, but memory usage and computation time increase significantly
Solution Approach 1:
The patent extracts only the essential feature (edge orientation) from the complex pixel data, rather than analyzing all pixel properties. By focusing specifically on edge orientation statistics and comparing these extracted features between current and historical boundaries, the system achieves differentiation with reduced computational overhead.
Solution Approach 2:
The patent changes the analysis parameter from raw pixel values or complex boundary strength metrics to edge orientation statistics. This parameter transformation simplifies the data representation while retaining the discriminative information needed to distinguish between abandoned and removed objects.
3Ease of operation
If threshold-based boundary strength comparison is used to classify events, then event classification is possible, but the threshold value is difficult to determine and may be scene-dependent
Solution Approach 1:
The system performs self-calibration by automatically learning the statistical distribution of edge orientations from historical boundary data. Instead of requiring manual threshold configuration, the system adapts to scene-specific characteristics through automated statistical analysis, making it universally applicable across different surveillance environments.
Solution Approach 2:
The patent incorporates feedback from historical boundary edge orientations to dynamically adjust classification decisions. By continuously learning from past data and comparing current edge statistics against historical patterns, the system automatically adapts to changing scene conditions without requiring manual intervention.
Data Source
AI summary
Disclosed is a computer-implemented method for classifying a detected region of change of a video frame as indicating an abandoned object or a removed object, the detected region of change is classified as an removed object if the temporal change of the edge consistency is from a more consistent state to a less consistent state, and the detected region of change is classified as an abandoned object if the temporal change of the edge consistency if from a less consistent to a more consistent state.