System and method for predicting mineral concentration

Abstract

The invention relates to the practical application of a machine learning technology in exploration of the underground of the earth and revealing of the possible position of an ore deposit. The present invention is directed to a method of predicting mineral concentration in a terrestrial area, the method comprising obtaining RF data elements in one or more polarizations from a radar mounted on a mobile platform, the RF data elements representing reflections of an RF scan from the terrestrial area; analyzing the RF data elements to produce an SAR data structure, where the SAR data structure comprises one or more polarization layers representing one or more polarizations, respectively, and where each polarization layer comprises a plurality of tile data elements representing a respective plurality of sub-regions of the scanned land region; and applying an ML model on the SAR data structure to predict values of mineral concentration buckets in at least one sub-region of the scanned land region.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of priority to U.S. Provisional Patent Application No. 63 / 394,186, filed on August 1, 2022, entitled “SYSTEM AND METHOD OF PREDICTINGMINERAL CONCENTRATION,” which is incorporated herein by reference as if fully set forth in its entirety herein. Technical Field

[0002] The present invention relates generally to geology. More specifically, the present invention relates to the practical application of machine learning and artificial intelligence techniques to exploring the Earth's subsurface and revealing the possible locations of mineral deposits using synthetic aperture radar (SAR). Background Art

[0003] As is well known, the problem of exploring the Earth's subsurface in search of mineral deposits has long been a subject of research. Today, a large number of methods have been developed for investigating the Earth's subsurface. These methods involve different types of drilling techniques, geological surveys, geochemical and geophysical exploration, and remote sensing.

[0004] Remote sensing is the process of detecting and monitoring the physical properties of an area at a distance, usually from a satellite, aircraft or drone. In the geological sciences, remote sensing is used as a complementary data acquisition method to support field observations, as it allows mapping of the geological properties of an area without the need for physical contact with the area being explored. "Sensing" is carried out via the detection of reflected electromagnetic radiation, which may be natural (e.g. by the sun) or artificial (e.g. by radar mounted on a satellite or aircraft).

[0005] Synthetic Aperture Radar (SAR) is considered one of the most advanced technologies for applied remote sensing. SAR exploits the motion of a radar antenna over a target area and the Doppler effect to provide finer spatial resolution than conventional stationary beam scanning radar. The distance traveled by a SAR device over a target during the period when the target scene is radiating creates a large synthetic antenna aperture (the size of the antenna), which in turn provides highly detailed analysis with a relatively small physical antenna.

[0006] Although SAR technology can potentially provide detailed and scalable investigation and exploration of mineral deposits at reasonable expense, in practice it has proven to be challenging. Even when detailed SAR analysis data for scanned land areas are available, the definition of the dependencies between specific SAR data features and the presence or absence of specific mineral deposits, let alone the possible concentrations of minerals, remains underdeveloped. As a result, the results of such analyses are not sufficiently valid and reliable. Summary of the invention

[0007] Therefore, there is a need for a system and method for predicting mineral concentrations in a particular land area that would improve the efficiency and reliability of applying SAR remote sensing technology for mineral deposit exploration purposes.

[0008] In order to overcome the shortcomings of the prior art, the following invention is provided.

[0009] In a general aspect, the present invention may be directed to a method for predicting mineral concentration in a land area by at least one processor, the method comprising obtaining radio frequency (RF) data elements in one or more polarizations from a radar mounted on a mobile platform, the data elements representing reflections of an RF scan from a land area; analyzing the RF data elements to generate a synthetic aperture radar (SAR) data structure, wherein the SAR data structure comprises one or more polarization layers, which respectively represent the one or more polarizations, and wherein each polarization layer comprises a plurality of patch data elements, which represent a corresponding plurality of sub-areas of the scanned land area; and applying a machine learning (ML) model on the SAR data structure to predict a range of values ​​of mineral concentration in at least one sub-area of ​​the scanned land area.

[0010] In another general aspect, the present invention may be directed to a system for predicting mineral concentrations, the system comprising: a non-transitory memory device in which an instruction code module is stored; and at least one processor associated with the memory device and configured to execute the instruction code module, whereby, executing the instruction code module, the at least one processor is configured to obtain radio frequency (RF) data elements in one or more polarizations from a radar mounted on a mobile platform, the radio frequency (RF) data elements representing reflections of an RF scan from a land area; analyze the RF data elements and generate a synthetic aperture radar (SAR) data structure, wherein the SAR data structure comprises one or more polarization layers, which respectively represent one or more polarizations, and wherein each polarization layer comprises a plurality of small block data elements, which represent corresponding plurality of sub-areas of the scanned land area; and apply a machine learning (ML) model on the SAR data structure to predict a value of a mineral concentration bin in at least one sub-area of ​​the scanned land area.

[0011] In some embodiments, each polarization layer includes data representing at least one of: (a) an amplitude of an RF scan reflection, and (b) a phase of the RF scan reflection.

[0012] In some embodiments, the method further includes computing one or more local incident angle values ​​based on the RF data elements, which represent the local incident angles of RF reflections from multiple sub-regions; generating a local incident angle map, which represents the computed local incident angle values; and further applying the ML model on the local incident angle map to predict the mineral concentration bucket values.

[0013] In some embodiments, the method further includes computing a digital elevation map based on the RF data elements, the digital elevation map representing elevations of a sub-area of ​​the scanned land area; and further applying the ML model on the digital elevation map to predict mineral concentration bucket values.

[0014] In some embodiments, the method further includes receiving spectral data elements representing a delineation of the scanned land area in at least one of an infrared (IR) band, a visible spectrum band, and an ultraviolet (UV) band; and further applying the ML model on the spectral data elements to predict the mineral concentration bucket values.

[0015] In some embodiments, the ML model includes at least one convolutional neural network (CNN) model comprising one or more input channels, each input channel configured to receive an input selected from: one or more polarization layers, a local incidence angle map, a digital elevation map, and a spectral data element.

[0016] In some embodiments, the ML model also includes at least one binary classifier model, which is suitable for receiving the output of the CNN model that pertains to at least one sub-region of the scanned land area; and calculating the probability that the mineral concentration in the at least one sub-region pertains to the concentration range defined by a specific mineral concentration bucket.

[0017] In some embodiments, the range of radio frequencies is selected from a list consisting of X-band, C-band, S-band, L-band, and P-band.

[0018] In some embodiments, one or more polarizations are selected from the following: (i) horizontal transmission-horizontal reception (HH) linear polarization, (ii) horizontal transmission-vertical reception (HV) linear polarization, (iii) vertical transmission-horizontal reception (VH) linear polarization, (iv) vertical transmission-vertical reception (VV) linear polarization, (v) right-hand transmission-right-hand reception (RR) circular polarization, (vi) right-hand transmission-left-hand reception (RL) circular polarization, (vii) left-hand transmission-right-hand reception (LR) circular polarization, and (viii) left-hand transmission-left-hand reception (LL) circular polarization.

[0019] In some embodiments, the method further includes: receiving a training dataset comprising at least one training SAR data structure corresponding to a training land area; and at least one annotation representing a bucket value of a mineral concentration in the training land area; and training an ML model based on the training dataset to predict a value of the mineral concentration in the training land area.

[0020] In some embodiments, the training data set also includes training local incidence angle maps corresponding to training land areas.

[0021] In some embodiments, the training data set also includes training digital elevation maps corresponding to training land areas.

[0022] In some embodiments, the training data set also includes training spectral data elements delineating at least a portion of a training land area. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portions of the specification. However, the invention as to its organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read in connection with the accompanying drawings, in which: Figure 1 is a block diagram depicting a computing device that may be included in a system for predicting mineral concentrations, according to some embodiments.

[0024] Figure 2A is a block diagram depicting a system for predicting mineral concentrations according to some embodiments; Figure 2B is a block diagram depicting aspects of training an ML model of a system for predicting mineral concentrations, according to some embodiments; and Figure 3 is a flow chart depicting a method of predicting mineral concentrations, according to some embodiments.

[0025] It should be understood that for simplicity and clarity of illustration, the elements shown in the figures are not necessarily drawn to scale. For example, for clarity, the size of some elements may be exaggerated relative to other elements. In addition, where deemed appropriate, reference numerals may be repeated in the figures to indicate corresponding or similar elements. DETAILED DESCRIPTION

[0026] Those skilled in the art will recognize that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the above-described embodiments should be considered in all respects as illustrative of the present invention as described herein rather than limiting. Therefore, the scope of the present invention is indicated by the appended claims rather than by the above description, and therefore, all changes within the equivalent meaning and scope of the claims are intended to be included therein.

[0027] In the detailed description below, many specific details are set forth to provide a thorough understanding of the present invention. However, it will be appreciated by those skilled in the art that the present invention can be practiced without these specific details. In other cases, in order not to obscure the present invention, well-known methods, procedures and components are not described in detail. Some features or elements described about one embodiment may be combined with features or elements described about other embodiments. For the sake of clarity, the discussion of the same or similar features or elements may not be repeated.

[0028] Although embodiments of the invention are not limited in this regard, discussions using terms such as, for example, "process," "compute," "calculate," "determine," "establish," "analyze," "examine," etc. may refer to (one or more) operations and / or processes of a computer, computing platform, computing system, or other electronic computing device that manipulate and / or transform data represented as physical (e.g., electronic) quantities within a computer's registers and / or memories into other data similarly represented as physical quantities within a computer's registers and / or memories or other information non-transitory storage media that may store instructions for performing the operations and / or processes.

[0029] Although embodiments of the present invention are not limited in this regard, the terms "plurality" and "a plurality" as used herein may include, for example, "multiple" or "two or more". The terms "plurality" or "a plurality" may be used throughout the specification to describe two or more components, devices, elements, units, parameters, etc. The term "set" when used herein may include one or more items.

[0030] Unless explicitly stated, the method embodiments described herein are not limited to a particular order or sequence. In addition, some of the described method embodiments or their elements may occur or be performed simultaneously, at the same time point or concurrently.

[0031] In the context of the description of the claimed invention, the term "concentration bucket" refers to a "range of concentration values," and thus, these terms may be used interchangeably.

[0032] In the context of the description of the claimed invention, "prediction of mineral concentrations in a land area" may refer to "prediction of mineral concentrations in the top layer and subsurface of a land area."

[0033] In some embodiments of the invention, the ML model may be an artificial neural network (ANN).

[0034] A neural network (NN) or artificial neural network (ANN), such as a neural network that implements machine learning (ML) or artificial intelligence (AI) functions, may refer to an information processing paradigm that may include nodes called neurons organized into layers with links between neurons. These links may pass signals between neurons and may be associated with weights. A NN may be configured or trained for a specific task, such as pattern recognition or classification. Training a NN for a specific task may involve adjusting these weights based on examples. Each neuron in an intermediate or final layer may receive an input signal, such as a weighted sum of output signals from other neurons, and may process the input signal using a linear or nonlinear function, such as an activation function. The results of the input and intermediate layers may be passed to other neurons, and the results of the output layer may be provided as the output of the NN. Typically, neurons and links within a NN are represented by mathematical constructs, such as activation functions and matrices of data elements and weights. A processor, such as a CPU or a graphics processing unit (GPU), or a dedicated hardware device may perform the relevant operations.

[0035] It should be apparent to those of ordinary skill in the art that various ML models may be implemented without departing from the essence of the present invention. It should also be understood that in some embodiments, the ML model may be a single ML model or a combination (integration) of ML models that, as a whole, implement the same functionality as a single ML model. Therefore, in view of the scope of the present invention, the above variations should be considered equivalent.

[0036] In some aspects, the following description of the claimed invention is provided in accordance with the task of revealing the location of lithium deposits (e.g., lithium carbonate, lithium oxide, etc.) or any other minerals (such as other metal deposits). Such specific examples are provided to make the description fully illustrative and are not intended to limit the scope of protection claimed by the invention. It should be understood by those of ordinary skill in the art that the embodiments of the claimed invention in accordance with such a task are provided as non-exclusive examples, and other actual embodiments may be covered by the claimed invention.

[0037] As is well known, artificial intelligence and machine learning techniques are very helpful in solving tasks where the connections and dependencies between input data and target output data are complex and uncertain (at least for humans) to be clearly defined. Therefore, the proposed invention combines a combination of specific ML and SAR techniques, as described in further detail herein. This combination is claimed to be effective for predicting mineral concentrations in a specific land area. This combination supports the realization of improved technical effects of increased efficiency and reliability in applying SAR remote sensing technology for mineral deposit exploration purposes. During the actual implementation of the present invention, the claimed technical effects have been additionally demonstrated.

[0038] Reference now Figure 1 , Figure 1 is a block diagram depicting a computing device that may be included in embodiments of a system for predicting mineral concentrations, according to some embodiments.

[0039] The computing device 1 may include a processor or controller 2, which may be, for example, a central processing unit (CPU) processor, a chip, or any suitable computing or operating device; an operating system 3; a memory device 4; an instruction code 5; a storage system 6; an input device 7, and an output device 8. The processor 2 (or one or more controllers or processors, which may span multiple units or devices) may be configured to implement the methods described herein, and / or to execute or act as various modules, units, etc. More than one computing device 1 may be included in a system according to an embodiment of the present invention, and one or more computing devices 1 may act as components of the system.

[0040] The operating system 3 may be or may include any code fragment (e.g., a code fragment similar to the instruction code 5 described herein) that is designed and / or configured to perform tasks involving coordination, scheduling, arbitration, supervision, control, or otherwise managing operations of the computing device 1 (e.g., scheduling the execution of software programs or tasks, or enabling software programs or other modules or units to communicate). The operating system 3 may be a commercial operating system. It should be noted that the operating system 3 may be an optional component, for example, in some embodiments, the system may include a computing device that does not require or include an operating system 3.

[0041] The memory device 4 may be or may include, for example, a random access memory (RAM), a read-only memory (ROM), a dynamic RAM (DRAM), a synchronous DRAM (SD-RAM), a double data rate (DDR) memory chip, a flash memory, a volatile memory, a non-volatile memory, a cache memory, a buffer, a short-term memory unit, a long-term memory unit, or other suitable memory unit or storage unit. The memory device 4 may be or may include a plurality of possibly different memory units. The memory device 4 may be a computer or processor non-transitory readable medium, or a computer non-transitory storage medium, such as RAM. In one embodiment, a non-transitory storage medium (such as a memory device 4, a hard drive, another storage device, etc.) may store instructions or codes that, when executed by a processor, may cause the processor to implement the method described herein.

[0042] The instruction code 5 may be any executable code, such as an application, program, process, task, or script. The instruction code 5 may be executed by the processor or controller 2, possibly under the control of the operating system 3. For example, the instruction code 5 may be an application that can predict mineral concentrations by applying an ML model on a SAR data structure, as further described herein. Although for clarity, Figure 1 A single instruction code 5 is shown in the figure, but the system according to some embodiments of the present invention may include multiple executable code fragments or modules similar to the instruction code 5, which can be loaded into the memory device 4 and cause the processor 2 to implement the method described herein.

[0043] The storage system 6 may be or may include, for example, a flash memory known in the art, a memory inside or embedded in a microcontroller or chip known in the art, a hard drive, a CD-Recordable (CD-R) drive, a Blu-ray Disc (BD), a Universal Serial Bus (USB) device, or other suitable removable and / or fixed storage unit. Various types of input and output data may be stored in the storage system 6 and may be loaded from the storage system 6 into the memory device 4 where it may be processed by the processor or controller 2. In some embodiments, the storage system 6 may be omitted. Figure 1 For example, the memory device 4 may be a non-volatile memory having the storage capacity of the storage system 6. Thus, although shown as a separate component, the storage system 6 may be embedded or included in the memory device 4.

[0044] The cache memory 9 may be or may include, for example, a level 1 (L1) cache module, a level 2 (L2) cache module, and / or a level 3 (L3) cache memory module as known in the art. The cache memory 9 may include, for example, an instruction cache memory space and / or a data cache memory space, and may be configured to cooperate with one or more processors (such as element 2) and / or one or more processing cores to perform at least one method according to an embodiment of the present invention. The cache memory 9 may typically be implemented on the same die or chip as the processor 2, and may therefore be characterized by a memory bandwidth that may be higher than the memory bandwidth of the memory device 4 and the storage system 6.

[0045] The input device 7 may be or may include any suitable input device, component or system, such as a detachable keyboard or keypad, a mouse, etc. The output device 8 may include one or more (possibly detachable) displays or monitors, speakers, and / or any other suitable output device. As shown in boxes 7 and 8, any applicable input / output (I / O) device may be connected to the computing device 1. For example, a wired or wireless network interface card (NIC), a universal serial bus (USB) device, or an external hard drive may be included in the input device 7 and / or the output device 8. It should be appreciated that, as shown in boxes 7 and 8, any suitable number of input devices 7 and output devices 8 may be operably connected to the computing device 1.

[0046] Systems according to some embodiments of the invention may include components such as, but not limited to, multiple central processing units (CPUs) or any other suitable general-purpose or specific processors or controllers (e.g., similar to element 2), multiple input units, multiple output units, multiple memory units, and multiple storage units.

[0047] Reference now Figure 2A , which depicts a system 100 for predicting mineral concentrations, according to some embodiments.

[0048] According to some embodiments of the present invention, the system 100 may be implemented as a software module, a hardware module, or any combination thereof. For example, the system 100 may be or may include a computing device such as Figure 1 Element 1. In addition, system 100 can be adapted to execute one or more instruction code modules (e.g., Figure 1 5) to request, receive, analyze, compute and generate various data in order to predict the values ​​of mineral concentration buckets in sub-areas of the scanned land area through the ML model 140, as further described in detail herein.

[0049] like Figure 2A and Figure 2B As shown, arrows may represent the flow of one or more data elements to and from system 100 and / or between modules or elements of system 100. For clarity, Figure 2A and Figure 2B Some arrows have been omitted.

[0050] In some embodiments, the system 100 may include a synthetic aperture radar (SAR) processing module 110. In some embodiments, the SAR processing module 110 may be configured to obtain radio frequency (RF) data elements 200A in one or more polarizations from a radar mounted on a mobile platform, which represent reflections of an RF scan from a land area. Alternatively, the RF data elements 200A may be provided by a third party input data provider (e.g., input data provider 200), for example, via a network connection.

[0051] As is known, in a non-limiting example, chemical compounds may include lithium, such as lithium carbonate, lithium oxide, etc., which have rotational absorption spectra in the microwave band. Therefore, in order to reliably detect these materials and quantify their abundance, in some embodiments, the range of radio frequencies for scanning is selected from a list consisting of X-band, C-band, S-band, L-band, and P-band.

[0052] In some embodiments, the SAR processing module 110 may be further configured to analyze each RF data element 200A to generate a SAR data structure 110A. To improve the reliability of the predictions further made by the ML model 140, the SAR processing module 110 may be configured to include one or more polarization layers 111A in the SAR data structure, which is also used as an input to the ML model 140, the polarization layers 111A respectively representing one or more polarizations. The SAR processing module 110 may be configured to select one or more polarizations from the following: (i) horizontal transmission-horizontal reception (HH) linear polarization, (ii) horizontal transmission-vertical reception (HV) linear polarization, (iii) vertical transmission-horizontal reception (VH) linear polarization, (iv) vertical transmission-vertical reception (VV) linear polarization, (v) right-hand transmission-right-hand reception (RR) circular polarization, (vi) right-hand transmission-left-hand reception (RL) circular polarization, (vii) left-hand transmission-right-hand reception (LR) circular polarization, (viii) left-hand transmission-left-hand reception (LL) circular polarization. The SAR processing module 110 may be configured to include a plurality of small-block data elements in each polarization layer 111A, which represent a corresponding plurality of sub-areas of the scanned land area.

[0053] Since the reflected wave function of the RF scan is usually complex, in order to make the further application of the ML model 140 effectively provide reliable predictions, the SAR processing module 110 can be configured to transform the wave function into an amplitude component and a phase difference component. Therefore, in some embodiments, the SAR processing module 110 can be configured to include data in each polarization layer 111A, which represents at least one of the amplitude of the RF scanned reflection and the phase of the RF scanned reflection. In some embodiments, the system 100 can be configured to use the polarization layer 111A including data representing the amplitude of the RF scanned reflection and the polarization layer 111A including data representing the phase of the RF scanned reflection as separate input channels of the ML model 140.

[0054] In order to perform the task of finding the location of the mineral deposit, various types of data can prove helpful. Therefore, the system 100 can be configured to perform data enhancement, that is, the system 100 can be configured to receive, operate and generate various types of data in order to further use these data as input channels of the ML model 140, which are either separate from the SAR data structure 110A or as additional data for each small piece of data element, respectively. In turn, the ML model 140 can also be configured to perform the function of data fusion (or cascading) (combining data from different input channels). Supplementing the input of the ML model 140 with such data can enhance the operation of the system 100, because these data can include deeply hidden features that can be highly correlated with the target output data (mineral concentration value prediction 100A). In addition, the use of a sufficiently sophisticated system 100 (and in particular the ML model 140) to fuse multiple data sources and data collection methods can improve the robustness of the system to any single insufficient aspects of these data sources, data types or collection methods. Therefore, the supply of additional input channels will amplify the technical effect of improving the efficiency of the operation of the ML model 140 and the reliability of the predictions of the ML model 140. Embodiments of a system 100 including such enhancements are described in further detail herein.

[0055] In some embodiments, system 100 may include a local incidence angle (LIA) processing module 120. In the context of the description of the claimed invention, the term "local incidence angle" refers to the angle between the ground normal at a particular location and the angle at which a satellite receives a reflected signal.

[0056] The LIA processing module 120 may be configured to obtain, for example, the RF data elements 200A provided by the input data provider 200 via a network connection. The LIA processing module 120 may also be configured to compute one or more local incidence angle values ​​based on the RF data elements 200A, which represent the local incidence angles of RF reflections from the plurality of sub-regions. The LIA processing module 120 may also be configured to generate a LIA graph 120A representing the computed local incidence angle values.

[0057] In some embodiments, the system 100 may include an elevation processing module 130. The elevation processing module 130 may be configured to obtain, for example, via a network connection, RF data elements 200A provided by an input data provider 200. The elevation processing module 130 may be configured to compute a digital elevation map 130A based on the RF data elements 200A, the digital elevation map 130A representing the elevation of a sub-area of ​​the scanned land area.

[0058] In addition, supplementary data belonging to different domains (e.g., spectral data, geological maps, etc.) can be used to improve the training of the ML model 140. Figure 2B Discuss further aspects of training.

[0059] Additionally or alternatively, in some embodiments, the system 100 may be configured to receive spectral data elements ( Figure 2A ), these data elements represent a depiction of the land area scanned in at least one of the infrared (IR) band, the visible spectrum band, and the ultraviolet (UV) band. The spectral data elements may be provided by a third party data provider.

[0060] The embodiments described herein provide a non-exclusive list of data types that may be used as input channels for the ML model 100. For example, additionally or alternatively, in some embodiments, the system 100 may be configured to receive a digital geological map representing fault lines in the earth and use the received geological map of the scanned land area as an additional input channel for the ML model 100.

[0061] In some embodiments, the system 100 may also be configured to apply an ML model 140 on the SAR data structure 110A, the LIA map 120A, and the digital elevation map 130A (and optional spectral data elements and geological map data elements) to predict values ​​of mineral concentration buckets in at least one sub-region of the scanned land area and generate concentration value predictions 100A.

[0062] In some embodiments, the ML model 140 may include a convolutional neural network (CNN) model (e.g., a CNN-based encoder 141). It should be understood that according to some embodiments of the present invention, the CNN model may be implemented as an encoder (e.g., encoder 141), and therefore, the terms "CNN model", "CNN-based model", "CNN-based encoder", and "encoder" may refer to the same element and may be used interchangeably herein. The encoder 141 may include one or more input channels, each of which is configured to receive input selected from one or more polarization layers 111A, LIA maps 120A, and digital elevation maps 130A, and additionally or alternatively, perform a cascade of the data received from different input channels. The encoder 141 may also be configured to perform multiple convolution operations on each input channel and produce a feature map (an efficient representation of the input data) as output data. The encoder 141 may include a convolution block that employs a batch normalization layer, which is found to provide higher accuracy than a maximum pooling layer. In addition, the system 100 can be configured to concatenate the output data of the encoder 141 with scalar features (e.g., orbital parameters) (e.g., whether the orbital direction is rising or falling relative to the North Pole, and whether the satellite's side-looking sensor is looking left or right during a scan). In some additional or alternative embodiments (not shown), multiple ML models 140 can be used, including multiple arrangements of ML models 140 (e.g., with different hyperparameter settings, different types of architectures, etc.), and the system 100 can use these scalar features (e.g., in a decision tree model) to select which encoder 141 to apply.

[0063] In some embodiments, the ML model 140 may include a single multi-class classifier model or a corresponding plurality of binary classifier models, such as a multi-layer perceptron (MLP) classifier 142. The MLP classifier 142 may be adapted to receive the output of the encoder 141 belonging to at least one sub-region of the scanned land area. The MLP classifier 142 may also be configured to compute a probability that the mineral concentration in the at least one sub-region belongs to a concentration range defined by a particular mineral concentration bucket (e.g., outputting a concentration value prediction 100A).

[0064] In some embodiments, the system 100 can be configured to predict average mineral concentrations (e.g., average lithium concentrations) for various quantities. For example, in some embodiments, the ML model 140 includes k - 1 MLP classifier 142, where k is the number of ranges (buckets), i.e. in the example provided k =7. Each MLP classifier 142 can be configured to predict whether the average lithium concentration in at least one sub-region (at the location of a given pixel) is above a specific preset value of a concentration bucket. The system 100 can be further configured to combine the outputs of the MLP classifiers 42 to achieve a resulting probability that a given sub-region belongs to each preset concentration bucket. For each MLP classifier 142, the concentration value can be calculated as follows: For the first bucket of MLP classifier 142: P 1 = 1 - Pr ( Target > V 1); For the MLP classifier 142 of the middle bucket: P i = Pr ( Target > V i )- Pr ( Target > V i-1 );as well as For the MLP classifier 142 of the last bucket: P k = Pr ( Target > V k-1 ), in P i is the calculated probability of whether the concentration value belongs to a preset concentration bucket of a specific MLP classifier 142, Pr is the concentration value V The probability of being less than a preset target value for a specific concentration bucket.

[0065] Furthermore, to arrive at the resulting concentration value prediction 100A, the system 100 may be configured to apply the estimated likelihood provided by each MLP classifier 142. For example, the system 100 may be configured to sort the outputs of the MLP classifiers 142 in descending order by their estimated likelihood and use the output of the MLP classifier 42 having the highest estimated likelihood as the resulting concentration value prediction 100A.

[0066] In some embodiments, system 100 may additionally be configured to perform training of ML model 140. Figure 2B Aspects of training the ML model 140 are further discussed.

[0067] Reference now Figure 2B, which depicts aspects of training ML model 140 of system 100 for predicting mineral concentrations, according to some embodiments.

[0068] In some embodiments, system 100 may be configured to receive all required training data from one or more training data providers (eg, training data provider 300 ) (described in more detail below).

[0069] In some embodiments, the training data may include input data samples 310A, supplementary data 320A (provided in additional or alternative embodiments) for penalizing the ML model 140, and output labels 330A (corresponding to the input data samples 310A). The input data samples 310A may include training SAR data structure samples 311A, training LIA map samples 312A, and training digital elevation map samples 313A.

[0070] The supplemental data 320A may include a spectral data element 321A and a geological map data element 322A. The geological map data element 322A may represent a digital geological map of earth fault lines, geological units, rock types, unit ages, and geomorphic features. The spectral data element 321A may represent a depiction of a land area scanned in at least one of an infrared (IR) band, a visible spectrum band, and an ultraviolet (UV) band.

[0071] Output labels 330A may include ground-truth concentration data 331A, which may include multiple annotations representing bucket values ​​of mineral concentrations in training land areas. The ground-truth concentration data 331A may be primarily created by measuring lithium concentrations in multiple drill holes at different depths ranging, for example, from 0.5 meters to 10 meters.

[0072] In some additional or alternative embodiments, the system 100 may include a training data preparation module (not shown). The training data preparation module may be configured to receive basic truth concentration data 331A.

[0073] The training data preparation module may also be configured to request the SAR processing module 110, the LIA processing module 120, and the elevation processing module 130 to provide input training data samples 310A, including SAR data structure samples 311A, training LIA map data samples 312A, and training digital elevation map data samples 313A, respectively, which correspond to training land areas (as indicated by the basic truth concentration data 331A). The SAR processing module 110, the LIA processing module 120, and the elevation processing module 130 may be configured to request corresponding RF data elements 200A corresponding to the training land areas from the input data supplier 200. The SAR processing module 110, the LIA processing module 120, and the elevation processing module 130 may also be configured to operate the training SAR data structure samples 311A, the training LIA map data samples 312A, and the training digital elevation map data samples 313A, respectively, based on the received RF data elements 200A.

[0074] In some embodiments, additionally or alternatively, the training data preparation module may be further configured to request the spectral data elements 321A and the geological map data elements 322A corresponding to the training land area from the training data supplier 300 and receive the requested data elements 321A and 322A.

[0075] The training data preparation module may also be configured to form a training data set including input data samples 310A corresponding to the training land areas, supplementary data 320A for penalizing the ML model 140 , and output labels 330A.

[0076] In some embodiments, forming a training data set may include the following procedures.

[0077] The training data preparation module can be configured to generate an image of the training land area based on the ground truth concentration data 331A, wherein the pixels are sampled into polygons of equal concentration buckets so that all of the image of the training land area is covered. The training data preparation module can be configured to perform a dilation operation around each pixel to form a plurality of square tile data elements representing a corresponding plurality of sub-regions of the scanned land area. In some embodiments, the training data preparation module can be configured to select and set the size of the tile data element.

[0078] The training data preparation module may also be configured to collate corresponding small blocks of data elements from each polarization layer 111A of the training SAR data structure samples 311A, and additionally or alternatively, to collate each of the training spectral data elements 321A, the training LIA map samples 312A, and the training digital elevation map samples 313A to form a training data set with corresponding image data channels. Thus, in the prepared training data set, each training sample is a small block of data elements (centered on the pixel to which the basic truth concentration data 331A belongs) comprising multi-channel image data. The training data preparation module may also be configured to sample the training data set using a stratified sampling technique in order to enforce equal representation of each class and thereby reduce overfitting. The training data preparation module may also be configured to split the training data set into a set of training samples and a set of validation samples (e.g., with a 75 / 25% bias toward training samples). In some embodiments, the training data preparation module may also be configured to exclude some sub-regions where ground truth labels are available for both training and validation sample sets in order to assess model generalizability and robustness to the distribution of input data.

[0079] In some embodiments, the system 100 may also include an ML training module 150 .

[0080] In some embodiments, the ML training module 150 can be configured to train the ML model 140 based on the training data set to predict the value of the mineral concentration in the training land area using the included multi-channel image data (e.g., the concentration value prediction 100A, such as Figure 2A as shown).

[0081] Since some target minerals (minerals whose concentrations can be predicted by the system 100, such as lithium) are rare earth elements (REEs), in most cases, there may be only a few examples of significant accumulation in the known topsoil. Therefore, it seems impossible to create a well-balanced dataset (e.g., a dataset with an equal or nearly equal number of labeled samples for each target class (e.g., a class corresponding to a specific concentration bucket)) for supervised learning of the classification task, which in turn makes it almost unattainable to train a reliable MLP classifier by using traditional supervised learning methods. On the other hand, there may be more SAR image data (e.g., SAR data structure sample 311A) of a certain land area (or a sub-area thereof) provided as input without considering the concentration of the target mineral in the area.

[0082] The present invention provides the following solution to the above problems: It is proposed herein to perform a two-stage training procedure, wherein the first stage (auxiliary training) may be for the training of the encoder 141 , and the second stage (downstream training) may be for the training of the MLP classifier 142 .

[0083] In some embodiments, the ML training module 150 may include a pretext training module 151 and a downstream training module 152 to perform the two-stage procedure of training the ML model 140 , as described in detail below.

[0084] The combined two-stage procedure may be referred to as self-supervised learning (SSL). The "auxiliary" training is an unsupervised stage, and it does not require any knowledge of mineral concentrations. Its main purpose is to make the ML model 140 "understand" the nature of the input data (e.g., it may be a concatenation of the SAR data structure sample 311A, the LIA map data sample 312A, and the digital assessment map data sample 313A), and to be able to construct an efficient representation 141A of the input data sample 310A, which may, for example, have a reduced dimensionality but focus on highly important input data features.

[0085] To do this, in some embodiments, the auxiliary training module 151 can be configured to: (i) concatenate the SAR data structure sample 311A, the LIA map data sample 312A, and the digital evaluation map data sample 313A to obtain the input data sample 310A, and (ii) train the autoencoder 140' model to reconstruct the input data sample 310A. The autoencoder 140' can have any architecture known in the art, for example, it can include an encoder 141 and a decoder 141' block, and it can be trained using ML methods that may be known to those skilled in the art. In some embodiments, the encoder 141 can be implemented as a model based on an encoding CNN. The term "encoder" should be understood in the broadest possible sense in this article, and the present invention is not limited to any specific architecture thereof.

[0086] After being sufficiently trained, the autoencoder 140' may be able to reliably reconstruct the input data sample 310A, and therefore, in the "information bottleneck" segment of the autoencoder 140' (the output of the encoder 141, which may also be referred to herein as a "representation"), may be able to compute a valid representation 141A of the input data sample 310A. When the training of the autoencoder 140' is considered complete, the decoder 141' portion may no longer be used, neither in further training nor in inference of the ML model 140.

[0087] In a second training phase (downstream training), the capabilities of the computationally efficient representation 141A developed in the first training phase may be used for further supervised training of the ML model 140 , in particular for training the MLP classifier 142 .

[0088] In some embodiments, the downstream training module 152 can be configured to receive input data samples 310A and output labels 330A, which can be based on ground truth concentration data 331A. The downstream training module 152 can also be configured to perform training of the ML model 140, including: (i) providing the input data samples 310A as input to the pre-trained encoder 141, (ii) computing, by the encoder 141, valid representations 141A of the input data samples 310A, and (iii) based on the valid representations 141A of the input data samples 310A and the corresponding ground truth concentration data 331A values ​​(e.g., each sample of the valid representation 141A can correspond to a certain land region (or a sub-region thereof) and be labeled by a corresponding ground truth concentration data 331A value measured in the region), further training the MLP classifier to predict values ​​of mineral concentrations in the training land regions (e.g., concentration value prediction 100A). Figure 2A The “downstream” training stage may involve supervised ML methods known in the art.

[0089] Thus, in some embodiments, the ML model 140 can be configured wherein an encoder 141 can receive input data (e.g., training input data samples 310A (if a training phase is involved); or inference input data samples in multi-channel or cascaded single-channel form, which include one or more polarization layers 111A, LIA maps 120A, and digital elevation maps 130A (if an inference phase is involved)); compute a valid representation 141A of the input data and pass it downstream to an MLP classifier 142, which can receive the representation 141A as input and compute a concentration value prediction 100A as output.

[0090] Therefore, the proposed configuration of the ML model 140 and the indicated two-stage procedure for its training can help improve the prior art by alleviating the problem of lack of substantial training data, thereby increasing the training process efficiency and, therefore, improving the reliability of mineral concentration value (or bucket) predictions.

[0091] Furthermore, in some additional or alternative embodiments, analysis of the supplemental data 320A may be involved to further improve the training of the ML model 140, particularly by serving as a basis for adjusting the penalty imposed on the ML model 140 for failing to provide a correct concentration prediction.

[0092] It is well known that the presence of certain mineral accumulations, such as lithium, can be explained as follows. The origin of lithium is volcanic activity, through which lithium-rich magma finds its way through the Earth's crust via fractures. Once on the Earth's surface, this lava is deposited as volcanic rocks such as pegmatites. The rock undergoes weathering and erosion, and its components are transported to accumulation basins, such as the Salars (where, for example, lithium accumulates in brine) or argillaceous clay deposits.

[0093] Thus, the following different phases can be detected: volcanic deposition, transport and accumulation. Each phase has an associated temporal and spatial scale that can be used to improve the training process.

[0094] For example, in some embodiments, the training of the ML model 140 can make predictions of mineral concentration values ​​(e.g., subsurface lithium concentrations in known accumulation areas) for the highest currently known spatial resolution based on the RF data elements 200A (e.g., on the order of L-band SAR image pixels (1-10 meters)). In addition, if the characteristic lengths of normal faults in geological maps are analyzed, which are a result of the spatial resolution of the underlying geophysical measurements (essentially 1-10 kilometers, the lowest spatial resolution), then areas of potential volcanic deposits can be detected. In addition, if the topography of the watershed / drainage system is analyzed (at a resolution of 10 meters to 1 kilometer), the paths of lithium transport from volcanic deposits to accumulation areas can be detected and evaluated.

[0095] Additionally or alternatively, the downstream training module 152 may be configured to receive supplemental data 320A for penalizing the ML model 140. The downstream training module 152 may also be configured to perform the following operations: (i) for each particular input data sample 310A, select a corresponding supplemental data 320A sample that represents the same training land region (or portion / sub-region thereof) at a desired resolution (as described above); (ii) analyze the spectral data elements 321A and the geological map data elements 322A to determine areas of potential volcanic deposits; (iii) analyze the geological map data elements and, additionally or alternatively, analyze the digital assessment map data elements of the sample 313A to detect potential pathways of lithium transport from the volcanic deposits; and (iv) determine areas of potential lithium accumulation in the training land region (or portion / sub-region thereof). The downstream training module 152 may also be configured to further perform training of the ML model 140 based on: (i) the detected areas of potential volcanic deposits, (ii) the detected pathways of potential lithium transport, and (iii) the determined areas of potential lithium accumulation. In particular, the downstream training module 152 can be configured to compute a loss function (e.g., a function of the difference between an estimate of lithium concentration (computed by the ML model 140 during training) and a true value (e.g., corresponding ground truth concentration data 331A)), and further adjust the value of the loss function based on at least one of: (i) detected regions of potential volcanic deposits, (ii) detected pathways of potential lithium transport, and (iii) determined regions of potential lithium accumulation. For example, the downstream training module 152 can more strongly penalize the ML model 140 for erroneously predicting relatively lower lithium concentration values ​​in sedimentary basin regions than would otherwise be the case.

[0096] In some embodiments, to detect areas of potential volcanic deposits, the downstream training module 152 may also be configured to use data from geological map data elements 322A, such as the major and minor rock types in each unit, the range of unit ages, and the proximity to normal faults, their length, age, and orientation (e.g., "strike and dip"). In locations where geological maps are not available, the required information may also be derived from geophysical measurements, including seismic reflectometry, magnetometry, and gravimetry. To detect potential lithium transport pathways via drainage systems, the downstream training module 152 may also be configured to compute derivative data from digital elevation model (DEM) samples 322A, including normalized elevation, slope, aspect, and curvature.

[0097] Thus, applying such supplemental data 320A can help the model 140 effectively rely on such areas that can be considered to indicate high surface soil lithium concentrations and which can be consistent with local transport pathways and further even consistent with the origin of volcanic deposits. Therefore, such refinement of the supplemental data 320A can further improve the training process and increase the reliability of mineral concentration value predictions (e.g., concentration value prediction 100A).

[0098] Additionally or alternatively, to penalize errors superlinearly with error magnitude, in some embodiments, the ML training module 150 may be configured to use cross entropy with focal loss with a gamma parameter of 2 as a loss function during training the ML model 140. To reduce problems caused by an unbalanced training dataset, the ML training module 150 may be configured to set the alpha parameter of the MLP classifier 142 equal to the inverse of the relative abundance of each class (corresponding to the desired output of the MLP classifier 142).

[0099] In some embodiments, the ML training module 150 may be configured to utilize an ADAM optimizer and / or Early Stopping and / or a linear learning rate scheduling technique when training the ML model 140 .

[0100] In some embodiments, during the training phase, the system 100 may include multiple permutations of the ML model 140. In some embodiments, the ML training module 150 may be configured to perform hyperparameter tuning of the detection threshold of each MLP classifier 142 in the multiple ML models 140 so as to produce optimal thresholds for recall and precision of each MLP classifier 142. The ML training module 150 may also be configured to perform a neural architecture search (NAS) to determine the architecture of the ML model 140 or its components (e.g., the encoder 141 and the MLP classifier 142) with maximum prediction performance, minimum memory consumption, minimum model size or inference time (i.e., the time required to obtain a prediction), etc. The ML training module 150 may also be configured to select a target ML model 140 from the multiple permutations of the ML model 140 based on the results of the NAS procedure.

[0101] Reference now Figure 3 , a flow chart is presented that depicts a method of predicting mineral concentrations by at least one processor according to some embodiments.

[0102] As shown in step S1005, at least one processor (eg, Figure 1The processor 2) may execute to obtain RF data elements 200A in one or more polarizations from a radar mounted on a mobile platform, the RF data elements 200A representing reflections of the RF scan from a land area. Step S1005 may be performed by the SAR processing module 110, the LIA processing module 120, and the elevation processing module 130 (see reference 100). Figure 2A described above).

[0103] As shown in step S1010, at least one processor (eg, Figure 1 The processor 2 of the SAR processing module 110 may perform analysis of the RF data element 200A to generate a SAR data structure 110A, wherein the SAR data structure 110A includes one or more polarization layers 111A, which respectively represent one or more polarizations, and wherein each polarization layer 111A includes a plurality of small block data elements, which represent a corresponding plurality of sub-areas of the scanned land area. Step S1010 may be performed by the SAR processing module 110 (as shown in reference Figure 2A described above).

[0104] As shown in step S1015, at least one processor (eg, Figure 1 The processor 2 of FIG. 10 may execute applying the ML model 140 on the SAR data structure 110A to predict the value of the mineral concentration bucket in at least one sub-region of the scanned land area (concentration value prediction 100A). Step S1015 may be implemented by the encoder 141 and the MLP classifier 142 (as shown in FIG. 10A ). Figure 2A described above).

[0105] As can be seen from the description provided, the claimed invention represents a system and method for predicting mineral concentrations in a specific land area. The claimed invention improves the efficiency and reliability of applying SAR remote sensing technology for mineral deposit exploration purposes.

[0106] Unless explicitly stated, the method embodiments described herein are not limited to a particular order or sequence. In addition, all formulas described herein are intended to be examples only, and other or different formulas may be used. Additionally, some of the described method embodiments or elements thereof may occur or be executed at the same time point.

[0107] Although certain features of the present invention have been illustrated and described herein, many modifications, substitutions, changes and equivalents may occur to those skilled in the art. Therefore, it should be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

[0108] Various embodiments have been presented. Of course, each of these embodiments may include features from other embodiments presented, and embodiments not specifically described may include various features described herein.

Claims

1. A method for predicting mineral concentrations in a land area by at least one processor, the method comprising: obtaining radio frequency data elements (RF data elements) in one or more polarizations from a radar mounted on a mobile platform, the RF data elements representing reflections of an RF scan from the land area; analyzing the RF data elements to generate a synthetic aperture radar (SAR) data structure, wherein the SAR data structure comprises one or more polarization layers respectively representing the one or more polarizations, and wherein each polarization layer comprises a plurality of tile data elements representing a respective plurality of sub-areas of the scanned land area; and A machine learning model (ML model) is applied on the SAR data structure to predict values ​​of mineral concentration buckets in at least one sub-area of ​​the scanned land area. 2 . The method of claim 1 , wherein each polarization layer includes data representing at least one of: (a) an amplitude of the RF scan reflection, and (b) a phase of the RF scan reflection.

3. The method according to any one of claims 1 and 2, further comprising: calculating one or more local angle of incidence values ​​based on the RF data elements, which represent local angles of incidence of the RF reflections from the plurality of sub-regions; generating a local incident angle map representing the calculated local incident angle values; and The ML model is further applied on the local incidence angle map to predict the mineral concentration bucket values.

4. The method according to any one of claims 1 to 3, further comprising: computing a digital elevation map representing elevations of a sub-area of ​​the scanned land area based on the RF data elements; as well as The ML model is further applied on the digital elevation map to predict the mineral concentration bucket values.

5. The method according to any one of claims 1 to 4, further comprising: receiving a spectral data element representing a delineation of a land area scanned in at least one of an infrared band (IR band), a visible spectrum band, and an ultraviolet band (UV band); as well as The ML model is further applied on the spectral data elements to predict the mineral concentration bucket values.

6. A method according to any one of claims 1-5, wherein the ML model comprises at least one convolutional neural network model, i.e., a CNN model, the CNN model comprising one or more input channels, each input channel being configured to receive an input selected from: the one or more polarization layers, the local incident angle map, the digital elevation map, and the spectral data elements.

7. The method according to any one of claims 1 to 6, wherein the ML model further comprises at least one binary classifier model adapted to: receiving an output of the CNN model belonging to at least one sub-region of the scanned land area; and A probability is calculated that the mineral concentration in the at least one sub-region falls within a concentration range defined by a particular mineral concentration bucket.

8. The method according to any one of claims 1-7, wherein the range of radio frequencies is selected from a list consisting of X-band, C-band, S-band, L-band and P-band.

9. The method according to any one of claims 1 to 8, wherein the one or more polarizations are selected from the following: (i) horizontal transmission-horizontal reception linear polarization, i.e., HH linear polarization, (ii) horizontal transmission-vertical reception linear polarization, i.e., HV linear polarization, (iii) vertical transmission-horizontal reception linear polarization, i.e., VH linear polarization, (iv) vertical transmission-vertical reception linear polarization, i.e., VV linear polarization, (v) right-hand transmission-right-hand reception circular polarization, i.e., RR circular polarization, (vi) right-hand transmission-left-hand reception circular polarization, i.e., RL circular polarization, (vii) left-hand transmission-right-hand reception circular polarization, i.e., LR circular polarization, and (viii) left-hand transmission-left-hand reception circular polarization, i.e., LL circular polarization.

10. The method according to any one of claims 1 to 9, further comprising: A training data set is received, wherein the training data set includes: at least one training SAR data structure corresponding to a training land area; and a plurality of annotations representing bucket values ​​of mineral concentrations in said training land region; and The ML model is trained based on the training data set to predict values ​​of mineral concentrations in the training land area.

11. The method according to any one of claims 1-10, wherein the training data set further comprises a training local incidence angle map corresponding to the training land area.

12. The method according to any one of claims 1-11, wherein the training data set further comprises a training digital elevation map corresponding to the training land area.

13. The method of any one of claims 1-12, wherein the training data set further comprises training spectral data elements delineating at least a portion of the training land area.

14. A system for predicting mineral concentration, the system comprising: a non-transitory memory device having instruction code modules stored therein; and at least one processor associated with the memory device and configured to execute the instruction code module, thereby executing the instruction code module, the at least one processor is configured to: obtaining radio frequency data elements (RF data elements) in one or more polarizations from a radar mounted on a mobile platform, the RF data elements representing reflections of an RF scan from a land area; analyzing the RF data elements and generating a synthetic aperture radar (SAR) data structure, wherein the SAR data structure comprises one or more polarization layers respectively representing the one or more polarizations, and wherein each polarization layer comprises a plurality of tile data elements representing a respective plurality of sub-areas of the scanned land area; and A machine learning model (ML model) is applied on the SAR data structure to predict values ​​of mineral concentration buckets in at least one sub-area of ​​the scanned land area.

Images