Handheld Metal Detection Using Vector Subspace Target Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing metal detectors struggle to accurately identify non-ferrous objects of interest when colocated with ferrous junk objects, as the spectral responses are often masked by the dominant responses from nearby metallic clutter, leading to misidentification and reduced detection capabilities.
Innovation Solution
The method involves forming a vector subspace based on multiple spectral response measurements at different frequencies or locations, and processing this subspace with target models to determine closeness, using techniques like singular value decomposition to identify the presence of targets by testing whether candidate spectral responses lie within this subspace.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional metal detectors measure spectral response at single or multiple frequencies, then they can identify metal targets, but they fail to accurately detect non-ferrous objects when colocated with ferrous junk objects due to dominant responses masking the target signals
Solution Approach 1:
The patent transforms the detection problem from frequency domain analysis to vector subspace analysis. Instead of analyzing spectral responses at individual frequencies, the system constructs a vector subspace from multiple spectral response measurements taken at different frequencies or locations. This dimensional transformation allows the detector to represent the complex spectral signature of colocated objects as a unified vector structure, enabling separation of target signals from dominant ferrous clutter through geometric relationships in the vector space rather than through frequency-by-frequency analysis.
Solution Approach 2:
The patent changes the fundamental parameter of analysis from frequency-specific spectral magnitude to vector subspace geometric properties. By representing spectral responses as vectors and analyzing their relationships through dot products, angles, and subspace projections, the system transforms the detection criterion from 'magnitude at frequency f' to 'geometric relationship in vector space'. This parameter change enables the detector to identify targets based on the angular relationship between measured spectral vectors and reference target model vectors, making the detection robust against dominant ferrous responses that would otherwise mask non-ferrous targets.
2Device complexity
If metal detectors use dominant spectral response for target identification, then processing is simpler, but non-ferrous targets obscured by ferrous junk are misidentified as the dominant ferrous objects
Solution Approach 1:
The patent introduces vector subspace as an intermediary representation between raw spectral measurements and target identification. Instead of directly comparing spectral magnitudes to identify targets, the system first transforms spectral responses into vector form and projects them onto a subspace spanned by reference target models. This intermediary vector subspace representation acts as a mediator that decouples the detection process from the dominance of ferrous responses, allowing non-ferrous targets to be identified through their unique vector geometric relationships rather than through direct spectral magnitude comparison.
Solution Approach 2:
The patent replaces traditional spectral analysis mechanisms with vector algebra operations. Instead of using frequency-domain filtering, magnitude thresholding, or spectral matching techniques, the system employs vector dot products, angle calculations, and subspace projections to identify targets. This substitution of mechanical signal processing with vector mathematical operations enables the detector to reliably identify non-ferrous targets by their vector geometric signature rather than by dominant spectral magnitude, fundamentally changing the detection mechanism from magnitude-based to geometry-based identification.
3Measurement precision
If multiple spectral response measurements are taken at different frequencies or locations, then better target identification is possible, but the data processing complexity and computational requirements increase
Solution Approach 1:
The patent segments the spectral response data into discrete measurement vectors, each representing the spectral signature at a specific frequency or location. By dividing the continuous spectral data into separable vector components, the system can apply linear algebra operations to construct the measurement matrix and compute the vector subspace. This segmentation transforms the complex continuous spectral analysis problem into a structured discrete vector space problem, making the processing of multiple frequency measurements more manageable through systematic matrix operations rather than continuous spectral analysis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively detects non-ferrous targets obscured by ferrous junk by approximating their spectral responses as linear combinations of multiple measurements, improving detection accuracy and reducing false positives.
Implementation Method 1
transmit electronics generating a repeating transmit signal cycle of a fundamental period, which is applied to an inductor, for example a transmit coil, which transmits a resulting varying magnetic field
Implementation Method 2
receive electronics that process a receive signal from a measured receive magnetic field
Data Source
AI summary
Provided is a method to detect at least one target using a handheld metal detector including transmitting a transmit magnetic field using a transmitter; receiving a receive magnetic field using a receiver to produce one or more receive signals; processing the one or more receive signals to produce a plurality of different processed signals; forming a representation of a vector subspace based on the plurality of different processed signals; processing the representation of the vector subspace with each of a plurality of target models individually to determine a closeness from each of the plurality of target models to the vector subspace; and producing an indicator output indicative of the presence of at least one target based on the closeness from each of the plurality of target models to the vector subspace.