A hand-held metal detector

Abstract

The utility model discloses a handheld metal detector, this handheld metal detector includes: the processing unit of user interface, bar piece and sensor head or coil shell, sensor head or coil shell installs on the bar piece, and the processing unit is connected to sensor head or coil shell through compound cable, and sensor head or coil shell built -in transmitter and receiver, wherein, transmitter emits emission magnetic field, receiver receives receiving magnetic field to produce one or more receiving signals, the processing unit processes one or more receiving signals to generate a plurality of different processing signals, and constructs the representation of vector sub -space based on a plurality of different processing signals, and the representation of vector sub -space is handled respectively with a plurality of target model each respectively, to determine the close degree between each target model and vector sub -space, and generates the indication output of indicating the existence of at least one target based on the close degree of each target model in a plurality of target models and vector sub -space.

Description

Technical Field

[0001] This utility model relates to the field of metal detection technology, and in particular to a handheld metal detector. Background Technology

[0002] Most metal detectors used for soil detection come in common forms: handheld battery-powered units, conveyor belt-mounted units, or vehicle-mounted units. Examples of handheld products include detectors for locating gold mines; detectors for explosive mines or ordnance; or detectors for coins and treasure. Examples of conveyor belt-mounted units include stray metal detectors used in ore mining operations; examples of vehicle-mounted units include devices for locating buried mines.

[0003] Such metal detectors typically (but not necessarily) include: transmitting electronics that generate a fundamentally recurring period of transmitting signals, which are applied to an inductor, such as a transmitting coil, thereby emitting a time-varying magnetic field, sometimes referred to as a transmitting magnetic field.

[0004] Such metal detectors may also include receiving electronics that, during one or more receiving cycles within a repetitive transmission signal cycle, process the received signal from a measured receiving magnetic field to generate an indicative output signal; the indicative output signal at least indicates the presence of at least one metallic target within the influence range of the transmission magnetic field.

[0005] During the processing of the received signal, the received signal may be sampled or demodulated to generate one or more target channels; the one or more target channels may be further processed to generate an indication output signal.

[0006] Treasure hunters typically target items located in geographical areas with historical traces of human habitation or activity. Therefore, these locations are often filled with various metallic debris or distracting objects.

[0007] In demining scenarios, landmines and improvised explosive devices are often found in areas that also contain a large number of metal debris or interfering objects. These objects may be related to conflict events or human habitation activities.

[0008] Metal detectors typically measure the complex spectral response of a target. This complex spectral response can be measured at a single frequency, characterized by amplitude and phase. It is also commonly measured at multiple frequencies or across multiple frequency bands. The measurement of the complex spectral response does not necessarily have to be described by amplitude and phase at different frequencies. An alternative is to use multi-channel response measurements, where different channels have selective sensitivity to different characteristics of the target's complex spectral response. For example, a multi-channel system with selective sensitivity to different frequencies can be used, in which case the target's spectral response can be expressed in the channel domain rather than the frequency domain. Based on the measurement results of the complex spectral response, the metal detector can generate an indication output that identifies the target.

[0009] The spectral responses produced by targets are diverse, depending on their shape, material, and spatial orientation. However, most targets tend to exhibit dominant characteristic spectral responses, which can serve as reliable identification criteria.

[0010] When multiple targets are densely distributed, a metal detector will simultaneously excite nearby targets, and the received spectral response is actually the superposition of the responses of each excited target. In this case, the received spectral response is usually the characteristic of the target that contributes the largest proportion to the response. For example, if a deeply buried non-ferrous coin produces a weak amplitude response, while a shallowly buried iron nail produces a strong amplitude response, the overall response is closer to the characteristic contributed by the iron nail than the characteristic of the non-ferrous coin.

[0011] The phenomenon of metallic debris obscuring the target's characteristic response is extremely common. Metal detectors often fail to accurately identify target features, rendering any excavation attempts pointless. In certain situations, the measured spectral response may be dominated by the response of co-located metallic debris, causing the metal detector to misjudge the presence of such debris at the location of the target of interest.

[0012] For treasure hunters and other metal detector operators, the ability to detect non-ferrous targets in environments containing ferrous debris is of paramount value. Metal detectors with enhanced non-ferrous target detection capabilities in such scenarios offer a significant advantage over other devices.

[0013] This invention provides an alternative solution that differs from existing technologies. Utility Model Content

[0014] According to a first aspect disclosed in this utility model, a handheld metal detector is provided, the handheld metal detector comprising:

[0015] A processing unit with a user interface;

[0016] rods;

[0017] A sensor head or coil housing; the sensor head or coil housing is mounted on the rod, the processing unit is connected to the sensor head or coil housing via a composite cable, and the sensor head or coil housing has a built-in transmitter and receiver;

[0018] The transmitter is used to emit a magnetic field;

[0019] The receiver is used to receive the receiving magnetic field to generate one or more receiving signals;

[0020] The processing unit is configured to process the one or more received signals to generate multiple different processing signals, and construct a representation of a vector subspace based on the multiple different processing signals; and to process the representation of the vector subspace with multiple target models respectively to determine the proximity between each target model and the vector subspace, and generate an indication output indicating the existence of at least one target based on the proximity between each target model and the vector subspace.

[0021] In one implementation, the processing unit processes the representation of the vector subspace with each of the multiple target models to determine the closeness between each target model and the vector subspace. The specific methods include:

[0022] Calculate the distance between each target model and a vector in the vector subspace.

[0023] In one implementation, the vectors in the vector subspace on which the distance is calculated are the vectors that are closest to each of the multiple target models.

[0024] In one implementation, the proximity of each of the plurality of target models to the vector subspace is determined by determining the orthogonality of each target model with respect to the vector subspace.

[0025] In one implementation, the proximity of each of the plurality of target models to the vector subspace is determined by determining the distance between each target model and the vector subspace.

[0026] In one implementation, the closeness of each of the plurality of target models to the vector subspace is determined by comparing the linear combination of the basis of the vector subspace with each target model.

[0027] In one embodiment, the specific way in which the processing unit processes the vector subspace representation includes: processing the complement space of the vector subspace representation.

[0028] In one implementation, the vector subspace includes at least two dimensions.

[0029] In one implementation, the plurality of different processed signals are generated based on received signals received at different times.

[0030] In one implementation, the plurality of different processed signals are generated based on received signals processed by different filters, each filter having a different impulse response.

[0031] In one implementation, the plurality of different processing signals are generated based on received signals received at different locations.

[0032] In one embodiment, the receiver includes at least two receiving coils, and each of the at least two receiving coils generates one of the one or more received signals.

[0033] In one embodiment, the transmitter includes at least two transmitting coils, and each of the at least two transmitting coils emits a different transmitting magnetic field.

[0034] In one embodiment, the specific method by which the processing unit constructs a vector subspace based on multiple different processing signals includes: constructing a vector subspace using a linear combination of the multiple different processing signals.

[0035] In one implementation, the vector subspace is a spectral representation of the space spanned by the transformed received signal.

[0036] In one implementation, the target model is a spectral representation of the target expected response after transformation.

[0037] In one implementation, the plurality of target models includes at least two target models characterizing soil response.

[0038] In one embodiment, the transmitting magnetic field is a multi-frequency magnetic field; the dimension of the vector subspace is equal to twice the frequency of the multi-frequency magnetic field.

[0039] In one implementation, the indication output is a digital target identification number, which indicates the identity information of at least one target.

[0040] In one implementation, the indication output includes at least two digital target identification numbers, which indicate the identity information of at least two targets and are displayed in two different areas of the user interface.

[0041] In one implementation, the indication output is an audio output that indicates the identity information of at least one target.

[0042] In one implementation, the indication output is an audio output that indicates the identity information of at least two targets.

[0043] In one implementation, the processing unit specifically generates an indication output based on the individual proximity between each target model characterizing the soil response and the vector subspace. Attached Figure Description

[0044] The embodiments of this disclosure will be described with reference to the accompanying drawings:

[0045] Figure 1 A handheld metal detector with a sensor head;

[0046] Figure 2 This demonstrates the general form of this disclosure;

[0047] Figure 3 The functional relationship between each component and the characteristic frequency is shown under three transmission frequencies according to one embodiment;

[0048] Figure 4 A model was presented to explain why traditional metal detectors struggle to accurately detect certain scenarios;

[0049] Figure 5 This shows a set of spectral candidate vectors corresponding to conductive circuits with a certain range of characteristic frequencies. value. Detailed Implementation

[0050] In the prior art, there are known methods to improve the ability of metal detectors to detect non-ferrous targets co-located with ferrous debris. Such methods typically attempt to separate the co-located target response in the metal detector response in the time domain, so that each independent response can be identified separately and without interference.

[0051] One approach involves using a detection coil with a smaller sensitivity coverage area, so that the coil primarily excites the first target without significantly exciting a nearby second target simultaneously. This effect can typically be achieved by miniaturizing the detection coil. However, this method reduces the detectable depth of deeply buried targets and decreases the effective coverage area of ​​a single scan.

[0052] Another approach is to apply only a slight low-frequency filter to the metal detector response (i.e., reduce the low-frequency filter strength). Low-frequency filtering can cause the response of the first target to blur and overlay the response of a neighboring second target, reducing the identification accuracy of the second target. Low-frequency filtering is typically used to suppress noise, especially power grid noise, and it usually improves the signal-to-noise ratio of metal detectors, thereby increasing the maximum detection depth. Therefore, reducing the low-frequency filter strength of the metal detector response will have the negative effect of reducing the potential detection depth.

[0053] Another approach is to effectively broaden the feature space region used to identify target objects. The principle is that even if the response does not perfectly match the expected response, as long as they are sufficiently similar within a certain tolerance region, the metal detector can still determine that a target object has been detected. Expanding the tolerance region will increase the true positive detection rate. However, a negative effect of this approach is that the false positive rate is expected to increase simultaneously, because the possibility of other objects being mistakenly identified as the target increases.

[0054] Alternative approaches attempt to model the metal detector response as a linear combination of a set of candidate responses. For example, US9239400B2 describes modeling the metal detector response as a linear combination of at least two candidate responses, which could be: ferrous / nonferrous target responses parameterized by characteristic frequencies, or soil responses. The optimal combination that best explains the measured response is found by searching the space of candidate response combinations, for example, by minimizing the sum of squared errors between the measured value and the optimal linear combination of candidate responses.

[0055] Modeling the metal detector response as a linear combination of candidate response sets has several major drawbacks:

[0056] First, the method relies on searching a set of candidate response combinations. Due to the combinatorial nature of candidate response combination searches, the search space grows exponentially. Modern metal detectors typically classify targets into approximately 100 different categories. If from... Select from 10 possible targets A combination of objectives exists. A number of possible combinations.

[0057] For example, if from Select from 10 possible targets When there is a target, there exists A number of possible combinations. If from Select from 10 possible targets When there is a target, there exists There are several possible combinations. The computational complexity of this combinatorial search quickly becomes unbearable in practical applications.

[0058] Secondly, although many targets can be accurately modeled as non-permeable conductive disks of varying radii and thicknesses, the number of candidate responses required to accurately characterize the diversity of massive amounts of debris is extremely large. Debris is typically ferrous and varies in shape and size, including irregular forms. It is known that the spectral response of many debris types, such as bottle caps and other ferrous objects, depends on their spatial orientation relative to the probe coil. The sheer size of the debris set further exacerbates the computational burden of searching for combinations of candidate responses.

[0059] This disclosure improves the ability of metal detectors to detect target objects in the presence of co-located unwanted clutter. In particular, this disclosure enhances the target detection capability of metal detectors in scenarios where the target object is adjacent to other metal objects or where the target object is obscured by the response of other objects that constitute noise or clutter.

[0060] Research has found that by utilizing multiple differentiated measurement information from multi-frequency spectral responses, such masked targets can be effectively detected.

[0061] Multiple differential measurements of the spectral response can be achieved in various ways, including: measurements by changing the position of the detection coil relative to the target location; measurements using multiple receivers at different spatial points at the target location; measurements using multiple transmitters at different spatial points at the target location; or combinations of the above measurement configurations. The key requirement is that the measurement configuration must ensure that the contribution amplitudes of the target object and nearby clutter exhibit a differential distribution in different measurement results.

[0062] All multiple differential measurement data tend to reside within a low-dimensional spectral subspace. The dimension of this subspace is approximately equal to the number of linearly independent spectral responses contributing to the measurement result. At least one measurement response is a linear combination of the contributing spectral responses. For example, if there are two targets, each contributing a single [to the measurement result]. If the spectral response is given, then all measurements tend to lie within a two-dimensional subspace, which is a subset of the larger n-dimensional spectral vector space. However, it should be noted that this conclusion is only approximately true due to noise.

[0063] The presence of a target object can be determined by detecting whether its expected spectral response approaches the low-dimensional spectral subspace. This disclosed method contrasts sharply with conventional methods, which determine which candidate spectral response a specific measurement result most closely approximates.

[0064] The essential difference between the two is that the present disclosure verifies whether a candidate response can be accurately approximated as a linear combination of multiple measurements, while the existing scheme verifies whether the measurement results can be accurately approximated as a single candidate response or a linear combination of multiple candidate responses.

[0065] Specifically, the spectral measurement vector is usually... The model is as follows:

[0066]

[0067] Where: K represents the number of independent spectral responses contributing to the measurement results. Represents the i-th spectral response. This represents the magnitude of the i-th response. This represents the noise component.

[0068] It is known that the spectral measurement vector can be used to... The target identity is estimated by comparing it with a dictionary of candidate spectral responses, i.e., testing which candidate spectral response best matches the measurement result. This requires considering that the candidate vectors may need scaling factors to match the measurements. For example, this can be achieved by selecting the best-fit candidate spectral response using the least squares method.

[0069] The method is effective in single-target scenarios, where the measurement results contain only a single dominant spectral response. However, it has limitations in multi-target scenarios, especially when the dominant response in the mixture originates from a non-target object. In such cases, existing solutions often misidentify the target as another object contributing to the dominant response in the mixture.

[0070] To improve this situation, this disclosure considers performing multiple measurements on the target mixture. Each measurement yields a spectral response measurement vector. . The set of measurements can be represented as a measurement matrix X (each measurement matrix X). As (column vectors), that is:

[0071]

[0072] in, The column vectors represent the independent n-dimensional spectral responses that constitute the measurement; The row vectors represent the contribution weights of each spectral response in the corresponding measurement; express 3D noise matrix

[0073] The core objective of this disclosure is to verify the spectral response of a specific candidate. Is it possible to be The column vector (or more generally, whether it is located in) (The columns are arranged within the space).

[0074] When there is no noise, if belong If the column space is a given, then it must belong to the column space. The column space. However, due to noise The existence of Usually at full rank, at this time all All belong to The column space. To suppress noise... The impact of column space estimation can be assessed using... The low-rank approximation is achieved. For example, for After performing singular value decomposition, the minimum singular value is set to zero.

[0075] This invention relates to evaluating the spectral response of specific candidate spectrometers. With measurement matrix column space (or The closeness of the low-rank approximation matrix (column space) to the candidate spectral response. When sufficiently close to the vector subspace, the candidate spectral response can be determined. The corresponding target exists. Therefore, this scheme can effectively detect target objects in the vicinity even when other clutter responses overlap in the time domain.

[0076] Figure 1 A handheld metal detector 1 is shown. The handheld metal detector 1 includes a processing unit 3 with a user interface. The processing unit 3 is connected via a composite cable 9 to a sensor head or coil housing 5 mounted on a rod 7. The sensor head or coil housing 5 houses a transmitter (which can also be understood as a transmitting winding) and a receiver (which can also be understood as a receiving winding). Figure 1 Not shown in the diagram. The composite cable 9 typically comprises multiple independent cables, for example: one transmitting cable for connecting the transmitter to the transmitting electronics within the processing unit 3, and another cable for connecting the receiver to the receiving electronics within the processing unit 3. The composite cable 9 may be replaced by a wireless connection between the sensor head or coil housing 5 and the processing unit 3.

[0077] Figure 2 A general form of this disclosure is shown. In this general form, a method based on... Figure 1 The handheld metal detector shown implements metal detection based on the following principles:

[0078] Step 101 involves emitting a transmitting magnetic field via a transmitter. Step 103 involves receiving a receiving magnetic field via a receiver to generate one or more receiving signals. These two steps can be implemented in various forms well known to those skilled in the art, with the core requirement being that the magnetic field emitted by the transmitter can excite a target, and the signal received by the receiver can indicate the presence of a target within the influence range of the emitted magnetic field. The specific form of the transmitter and receiver varies depending on the design and implementation of the metal detector.

[0079] The next step 105 includes processing the one or more received signals to generate multiple different processed signals. The methods for obtaining the differentiated processed signals are varied: in one embodiment, the different processed signals are generated based on one or more received signals received at different times. In another embodiment, the multiple processed signals are generated based on received signals received at different locations. For example, the receiver may include at least two receiving coils, each coil generating one or more received signals. In one embodiment, the multiple processed signals are generated based on received signals processed by different filters, each filter having a different impulse response.

[0080] Step 107 includes constructing a representation of a vector subspace based on multiple different processed signals. In one embodiment, the vector subspace includes at least two dimensions. In one embodiment, the vector subspace is a spectral representation of the space generated after the received signal is transformed. In one embodiment, the transmitted magnetic field is a multi-frequency magnetic field, and the number of dimensions of the vector subspace is equal to twice the number of frequencies of the multi-frequency magnetic field. In one embodiment, constructing the vector subspace based on multiple processed signals includes constructing the subspace using a linear combination of the processed signals.

[0081] Next step 109 involves processing the representation of the vector subspace with each of the multiple target models to determine the proximity between each target model and the vector subspace. As an example, this can be achieved by calculating the distance between each target model and a vector in the vector subspace. The vector in the vector subspace used to calculate the distance is the vector closest to each of the multiple target models. The proximity between each target model and the vector subspace can be determined by determining the degree of orthogonality of each target model with respect to the vector subspace; alternatively, the proximity between each target model and the vector subspace can be determined by determining the distance between each target model and the vector subspace; or alternatively, the proximity between each target model and the vector subspace can be determined by comparing a linear combination of the basis of the vector subspace with each target model.

[0082] In one implementation, the step of processing the vector subspace representation includes processing the complement space of the vector subspace.

[0083] In one implementation, the target model is a spectral representation of the expected target response after transformation. In one implementation, the target model includes at least two models characterizing the soil response.

[0084] The next step 111 involves generating an indication output that indicates the target based on the proximity of each target model among the plurality of target models to the vector subspace. The step of generating the indication output may depend on the individual proximity between each target model characterizing the soil response and the vector subspace.

[0085] The signal processing portion of the method and its various implementations can be entirely implemented in software. The software, including various instructions, can be stored in a non-transitory computer-readable medium within the metal detector. The signal processing flow will be described in detail below.

[0086] In the first embodiment, a three-frequency metal detector is provided. The received signal is demodulated to generate three in-phase and quadrature components at different frequencies, thereby enabling detection at time... Generate measurement vectors:

[0087]

[0088] in: and Representing time respectively No. In-phase and quadrature components of the frequency.

[0089] exist Multiple measurement data were collected at different times to form a measurement matrix. :

[0090]

[0091] Assuming at most exist Different targets for measurement data Contribution response. For noise reduction processing, generate... Low-rank approximation matrix: Assume

[0092]

[0093] for Singular value decomposition (where , It is an orthogonal matrix. for A rectangular diagonal matrix with singular values ​​satisfying Construct the rank-C approximation matrix. :

[0094]

[0095] in and They are respectively , The The equivalent representation is:

[0096]

[0097] in for A rectangular diagonal matrix whose singular values ​​satisfy... ,forward One diagonal element is a non-zero singular value, and the rest of the elements are zero. As The low-rank approximation, The vector subspace spanned by the column vectors is a subset of the larger spectral vector space.

[0098] In subsequent analysis, consider The Moore-Penrose pseudoinverse will be of significant value. The pseudoinverse can be represented by singular value decomposition as:

[0099]

[0100] in for The pseudo-inverse matrix is ​​constructed by transposing the original matrix and then replacing the non-zero diagonal elements with their multiplicative inverses.

[0101] Further examine the following products:

[0102]

[0103] in for Diagonal matrix, its front One diagonal element is 1, and the rest are 0.

[0104] Subsequently, candidate vectors for specific spectra were verified. Is it possible to measure the data? Make a contribution. If Contribution to Then the expectation exists. A certain linear combination Make:

[0105]

[0106] In other words, expectations lie in Within the vector subspace spanned by column vectors. This can be assessed through... and The degree of closeness is used to measure the likelihood of the expectation being true, where It is the optimal fit estimate, which makes and Minimize the Euclidean distance between them. Geometrically, this can be understood as: evaluating... and Zhang Cheng's Euclidean distance closeness to quantum space. Evaluation and There are various methods for determining the proximity of Zhang Chengzi's subspace: other distance metrics can be used, such as... Manhattan distance in norm Chebyshev distance Alternatively, angular distance metrics, such as cosine similarity, can be used.

[0107] Optimal fit estimate The formula for calculation is:

[0108]

[0109] symbol This indicates the Moore-Penrose pseudo-inverse.

[0110] residual vector The formula for calculation is:

[0111]

[0112] We applied the conclusions from the previous section. .

[0113] Sum of Squares of Residuals It is a measure Can it be precisely represented as The index of the linear combination of column vectors. The residual sum of squares. This can be expressed as:

[0114] Theoretical predictions indicate that, under noise-free conditions, if Those who contribute to X do not exceed One of the independent spectral responses, then ;otherwise, .

[0115] Therefore, a set of candidate spectral responses can be tested, and based on To determine whether a specific response contributes to the measurement data, for example when... ( A contribution is determined when a preset threshold is met. According to this mechanism, if both candidate spectral responses contribute to the measurement data, then the sum of squared residuals of both should satisfy... Therefore, we can... Or a similar criterion, to detect the presence of a specific spectral response in a mixed spectral response.

[0116] In the second implementation, the spectral candidate response After normalization, the sum of squared residuals can now be expressed as:

[0117]

[0118] in for Diagonal matrix, its front One diagonal element is 1, and the rest are 0.

[0119] In the third implementation, an alternative objective function is constructed. :

[0120]

[0121] Note: Minimize of Simultaneously maximize Therefore, it can be made possible by selection. The candidate response with the largest value is evaluated among a set of candidate responses. Most likely to exist.

[0122] Consider defining the characteristic frequency. Conductive circuits at three frequencies ( , , Given the in-phase and quadrature responses ( and Spectral candidate vectors:

[0123]

[0124] in,

[0125]

[0126]

[0127] For example, Figure 3 This illustrates the functional relationship between the components and the characteristic frequency at three transmission frequencies in one implementation. In this example, Figure 3 Demonstrated at the transmission frequency , , conditions, Each component varies with the characteristic frequency The curves showing the changes. Specifically, traces 19 and 25 represent... The in-phase and quadrature responses; traces 21 and 27 respectively represent The in-phase and quadrature responses; traces 23 and 29 respectively represent The in-phase response and quadrature response.

[0128] Further consider the definition of the measurement matrix X: its column vectors consist of two spectral candidate vectors. and The linear combination constitutes, that is:

[0129]

[0130] Where row vector and This represents the magnitude of the candidate spectral vector in each measurement.

[0131] In this example, such as Figure 4 As shown, The amplitude of (trace 31) was deliberately set to be much greater than (Trace 33). The scenario simulates the following situation: by The coin response characterized by the component is compared to that represented by The component characterization of aluminum foil impurities results in a weak response. In this case, conventional metal detectors struggle to accurately identify coins because the overall response in all measurements is dominated by the aluminum foil impurity response.

[0132] Figure 5 Trajectory 35 illustrates a set of candidate spectral vectors corresponding to a set of characteristic frequency conductive loops. Values ​​(as mentioned earlier, based on the measurement matrix) (Rank 4 approximation). It can be clearly observed that the pseudo-spectrum has two significant peaks at approximately 10 kHz and 100 kHz, corresponding to the two spectral vectors contributing to the measurement matrix. Therefore, even with strong interference responses from clutter, the presence of a target at 10 kHz can still be detected.

[0133] In the fourth embodiment, the target object to be identified is soil containing metallic impurities. That is, when the measurement data includes a mixture of soil spectral responses and metallic impurity spectral responses, we aim to estimate the soil spectral response.

[0134] Metal detectors typically employ a calibration technique called "ground balancing," which adjusts signal processing parameters to make the device unresponsive to soil. This can be achieved in various ways: usually by pre-setting or measuring the soil's spectral response, and then suppressing components in the metal detector's response that are in the same direction as the spectral response vector, thus rendering the device insensitive to soil.

[0135] Measuring the spectral response of soil in environments contaminated with metallic debris is challenging. Typically, operators must locate metal-free areas and perform a "ground equilibration" process to obtain soil spectral response measurements. This is impractical in many scenarios: for example, finding clean areas is difficult and time-consuming when there is a large amount of metallic debris in the environment. Furthermore, soil response measurements may be contaminated by responses from adjacent metallic debris, reducing the effectiveness of the "ground equilibration" process. The techniques disclosed herein can be used to estimate the soil spectral response in the presence of metallic debris. In this case, the spectral response is calculated for a set of corresponding candidate soil spectral responses. Calculate the previously defined Indicators. The soil spectral response can be estimated as... Maximize the spectral response of a specific candidate soil.

[0136] In this specification, the terms "surface" and "soil" are used interchangeably. Those skilled in the art will understand that these terms refer to the surface material that may bury the target object. This surface layer is typically solid, may be homogeneous or a mixture of various soil types, and may contain moisture or water.

[0137] Those skilled in the art should understand that information and signals can be represented using a variety of technical means. For example, the data, instructions, commands, information, signals, bits, symbols, and chips mentioned above can be expressed through voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or light particles, or any combination thereof.

[0138] In this specification and subsequent claims, unless the context otherwise requires, the terms “comprising,” “including,” and variations thereof such as “including” and “including” shall be understood to include the stated integers or groups of integers, but not to exclude any other integers or groups of integers.

[0139] Any reference to prior art in this specification does not constitute and should not be construed as implying any form of confirmation that the prior art constitutes common general knowledge in the art.

[0140] Those skilled in the art should understand that the scope of this disclosure is not limited to the specific application scenarios described herein, nor is it limited to the preferred embodiments of the specific elements and / or features described or illustrated herein. It should be understood that this disclosure is not limited to the disclosed embodiments, but allows for numerous rearrangements, modifications, and substitutions while adhering to the scope of protection defined by the following claims.

[0141] It should be noted that the following claims are provisional claims only, provided as a model of claims that may be asserted, and are not intended to limit the scope of claims in any future patent application that may be filed based on this application. Technical features may be added to or omitted from the example claims to further define or redefine the scope of protection.

Claims

1. A handheld metal detector, characterized in that, The handheld metal detector includes: A processing unit with a user interface; rods; A sensor head or coil housing; the sensor head or coil housing is mounted on the rod, the processing unit is connected to the sensor head or coil housing via a composite cable, and the sensor head or coil housing has a built-in transmitter and receiver; The transmitter is used to emit a magnetic field; The receiver is used to receive the receiving magnetic field to generate one or more receiving signals; The processing unit is configured to process the one or more received signals to generate multiple different processing signals, and construct a representation of a vector subspace based on the multiple different processing signals; and to process the representation of the vector subspace with multiple target models respectively to determine the proximity between each target model and the vector subspace, and generate an indication output indicating the existence of at least one target based on the proximity between each target model and the vector subspace.

2. The handheld metal detector according to claim 1, characterized in that, The processing unit processes the representation of the vector subspace with each of the multiple target models to determine the closeness between each target model and the vector subspace. The specific methods include: Calculate the distance between each target model and a vector in the vector subspace.

3. The handheld metal detector according to claim 1, characterized in that, The proximity of each of the multiple target models to the vector subspace is determined by determining the orthogonality of each target model relative to the vector subspace.

4. The handheld metal detector according to claim 1, characterized in that, The proximity of each target model in the plurality of target models to the vector subspace is determined by determining the distance between each target model and the vector subspace.

5. The handheld metal detector according to claim 1, characterized in that, The closeness of each target model to the vector subspace is determined by comparing the linear combination of the basis of the vector subspace with each target model.

6. The handheld metal detector according to claim 1, characterized in that, The specific way the processing unit processes the vector subspace representation includes processing the complement space of the vector subspace representation.

7. The handheld metal detector according to claim 1, characterized in that, The multiple different processed signals are generated based on received signals received at different times.

8. The handheld metal detector according to claim 1, characterized in that, The multiple different processed signals are generated based on received signals received at different locations.

9. The handheld metal detector according to claim 1, characterized in that, The receiver includes at least two receiving coils, and each of the at least two receiving coils generates one of the one or more received signals.

10. The handheld metal detector according to claim 1, characterized in that, The transmitter includes at least two transmitting coils, and each of the at least two transmitting coils emits a different transmitting magnetic field.

11. The handheld metal detector according to claim 1, characterized in that, The specific methods by which the processing unit constructs a vector subspace based on multiple different processed signals include: A vector subspace is constructed using a linear combination of the multiple different processed signals.

12. The handheld metal detector according to claim 1, characterized in that, The vector subspace is the spectral representation of the space spanned by the transformed received signal.

13. The handheld metal detector according to claim 1, characterized in that, The target model is the spectral representation of the target expected response after transformation.

14. The handheld metal detector according to claim 1, characterized in that, The multiple target models include at least two target models characterizing soil response.

15. The handheld metal detector according to claim 1, characterized in that, The transmitting magnetic field is a multi-frequency magnetic field; the dimension of the vector subspace is equal to twice the frequency of the multi-frequency magnetic field.

16. The handheld metal detector according to claim 1, characterized in that, The indication output includes at least two digital target identification numbers, which indicate the identity information of at least two targets and are displayed in two different areas of the user interface.

17. The handheld metal detector according to claim 1, characterized in that, The indicated output is an audio output that indicates the identity information of at least two targets.

18. The handheld metal detector according to claim 14, characterized in that, Specifically, the processing unit generates an indication output based on the individual proximity between each target model representing the soil response and the vector subspace.

Images