Histogram-Based Compound Object Splitting for Security Scanners
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Volumetric imaging devices often incorrectly group multiple objects as a single compound object, leading to inaccurate threat detection in security scanners, as the properties of the compound object cannot be effectively compared to known threat or non-threat items, resulting in unnecessary inspections or missed threats.
Innovation Solution
A histogram-based compound object splitting technique is employed to identify and split potential compound objects into sub-objects by analyzing voxel feature value distributions, assigning sub-object labels, and refining boundaries to distinguish separate objects within the imaging data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a segmentation algorithm groups multiple objects as a single compound object, then the segmentation process is simplified and faster, but the threat detection accuracy deteriorates because the properties of the compound object cannot be effectively compared with known threat and non-threat items
Solution Approach 1:
The patent applies segmentation by dividing the compound object into separate sub-objects based on density thresholds. The segmentation algorithm analyzes the density distribution of voxels within a detected object and splits it into multiple sub-objects when distinct density peaks are identified, allowing each sub-object to be evaluated independently for threat detection while maintaining processing efficiency
2Ease of operation
If compound objects are treated as single units, then the inspection process is streamlined, but false positives increase because the diluted properties of compound objects resemble non-threat items
Solution Approach 1:
The system segments compound objects into sub-objects with distinct density characteristics. By separating objects with different densities (e.g., liquids and solids), the system can identify sub-objects that match threat profiles even when the original compound object would have been misclassified, thereby reducing false positives while maintaining operational simplicity
Solution Approach 2:
The patent applies local quality by analyzing different regions of the compound object based on their density properties. Each sub-object is evaluated based on its local density characteristics rather than the averaged properties of the entire compound object, allowing accurate identification of threat-containing sub-objects without being diluted by non-threat components
3Loss of time
If compound objects are treated as single units, then the processing time is reduced, but threat detection accuracy deteriorates because the diluted properties of compound objects resemble non-threat items
Solution Approach 1:
The system performs rapid segmentation by using density-based thresholds to split compound objects. The algorithm efficiently identifies density peaks and divides the object into sub-objects in a computationally efficient manner, maintaining fast processing speeds while improving threat detection accuracy through the analysis of individual sub-object properties
Data Source
AI summary
Certain imaging systems, such as automatic explosives detection systems, employ techniques that utilize image processing, feature extraction and decision making steps to detect threats in images. Such techniques use segmentation as a first algorithmic step, which extracts data representing objects from image data. Some of the extracted objects are actually composed of multiple distinct physical objects. For these compound objects discrimination becomes difficult because computed object properties are less specific than properties computed for a single physical object. A technique is described which includes splitting such compound objects by separating the data of each component from the rest of the data and using properties of density histograms based on voxel distributions in both density and spatial domains.