WO2012063241A1 - Système et procédé de détection de champs de mines - Google Patents

Système et procédé de détection de champs de mines Download PDF

Info

Publication number
WO2012063241A1
WO2012063241A1 PCT/IL2011/000872 IL2011000872W WO2012063241A1 WO 2012063241 A1 WO2012063241 A1 WO 2012063241A1 IL 2011000872 W IL2011000872 W IL 2011000872W WO 2012063241 A1 WO2012063241 A1 WO 2012063241A1
Authority
WO
WIPO (PCT)
Prior art keywords
spectral
roi
minefield
image
parameters
Prior art date
Application number
PCT/IL2011/000872
Other languages
English (en)
Inventor
Avi Buzaglo Yoresh
Original Assignee
Avi Buzaglo Yoresh
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Avi Buzaglo Yoresh filed Critical Avi Buzaglo Yoresh
Publication of WO2012063241A1 publication Critical patent/WO2012063241A1/fr
Priority to ZA2013/03377A priority Critical patent/ZA201303377B/en

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry

Definitions

  • the present invention generally relates to system and method for minefield identification, more specifically, the present invention relates to system and method for minefield identification by remote sensing of exposed chemicals.
  • US 7157714 teaches a thermal imaging method to detect subsurface objects, but fails to show how biological processes may contribute to the speed and efficiency of minefield detection over large areas. Another method finds buried mines by employing plants sensitive to presence of TNT in the soil. It suggests genetically manipulating plants so that they may change their behavior in presence of TNT, for example changing color. US application 20050289662, for example, teaches how this can be achieved in detail, but fails to show how remote sensing can be employed to detect changes in vegetation in a general case. GB 1046437 teaches collecting particles carries in the air by a flying platform and analyzing them therein.
  • Remote sensing sensors such as IR, multispectral and color visible are used for the detection of mines and minefields primarily for airborne or forward looking vehicles platforms.
  • Most of the ongoing research projects on airborne minefields detection are military programs. Among them the UK initiative REMDIS (Bishop, 1998), the US program ASTAMIDS (Simard, 1998) and German BTG study (Scheerer, 1996) were published in press. All these programs are aimed at detecting mines using optronic sensors. At this stage the results for direct mine detection (surface or buried) have been demonstrated foe Anti-Tank mines by observing the thermal contrast or color contrast between the perturbed soil and the surroundings.
  • Spectroscopy is the science of measuring the spectral distribution of photon energies associated with radiation that may be transmitted, reflected, emitted, or absorbed upon passing from one medium to material objects.
  • Imaging spectroscopy is the special case in which spectral characteristics and variations of one variable tie to two additional variables, to generate color composite images, ratio and principal-components images, and classification maps.
  • images that represent the effects of diagnostic absorption bands can be produced to show specific spatial variability of certain material features that one or more such bands discretely identify.
  • Hyperspectral imaging is a technique that combines both conventional imaging and spectroscopy. Using this technology, both the spatial and spectral information of an object can be acquired.
  • the imaging produces 3D images or Hyperspectral image cubes and uses optical elements, lenses, spatial filters and image sensors to capture the content at multiple wavelengths.
  • Hyperspectral sensors are expected to find applications in reconnaissance and surveillance missions. These techniques can be used to get more information than traditional broadband cameras. Additionally, since these sensors operate over several wavelength bands, variations due to changes in physical parameters can be easily discovered. These reconnaissance missions can be used for mine detection and minefield delineation.
  • the system comprises:
  • a sensing unit for receiving the electromagnetic radiation of the ROI and generating at least one image of the electromagnetic radiation of the ROI, the ROI being characterized by a specific geographical location;
  • an analyzing unit adapted to classify the probability of the at least one image to be a minefield by using one or more spectral parameters associated with at least one exposed chemical of the ROI;
  • the spectral parameters are selected from the group consisting of: SR, NDVI, PVI, SAVI, NLI, RDVI, MSR, WDVI, MNLI, NDV SR, SAVI*SR, TSAVI, green-ratio, Red-NIR parameters, NIR-SWIR1 slope, SAM algorithm in SWIR2, PPI, Nitrogen spectral index, and any combination thereof.
  • MSSD minefield spectral signature dataset
  • MSSF minefield spectral signature function
  • CGST control group signature function
  • spectral parameters are vegetation parameters. It is another object of the present invention to disclose a system as defined, wherein the spectral parameters are grouped into categories selected from the group consisting of: broadband greenness, narrowband greenness, light use efficiency, canopy nitrogen, dry carbon, senescent carbon, leaf pigments, canopy water content, and any combination thereof.
  • the sensing unit comprising a sensor selected from the group consisting of: ASTER sensor, AISA-Dual sensor, HyMap sensor.
  • sensing unit is connectable and transpprtable by a transportation object selected from the group consisting of: aircraft, satellite, ground vehicle, and any combination thereof.
  • sensing unit further comprising at least one of: a satellite, an airborne camera, a camera carries by a land vehicle, and any combination thereof.
  • a classification technique selected from the group consisting of: K-mean, ISODATA, Spectral Angle Mapper, Artificial Neural Network, Pattern recognition, genetic algorithm, and any combination thereof.
  • the method comprises steps of:
  • spectral parameters are selected from the group consisting of: SR, NDVI, PVI, SAVI, NLI, RDVI, MSR, WDVI, MNLI, NDVPSR, SAV SR, TSAVI, green-ratio, Red-NIR parameters, NIR-SWIR1 slope, SAM algorithm in SWIR2, PPI, Nitrogen spectral index, and any combination thereof.
  • MSSF minefield spectral signature function
  • CGST control group signature function
  • spectral parameters are vegetation parameters. It is another object of the present invention to disclose a method as defined, further comprising step of selecting the spectral parameters from categories consisting of: broadband greenness, narrowband greenness, light use efficiency, canopy nitrogen, dry carbon, senescent carbon, leaf pigments, canopy water content, and any combination thereof.
  • a classification technique selected from the group consisting of: K-mean, ISODATA, Spectral Angle Mapper, Artificial Neural Network, Pattern recognition, genetic algorithm, and any combination thereof.
  • Fig. la schematically illustrates the context in which the present invention is employed
  • Fig. lb schematically illustrates the algorithm according to which a specific embodiment of the present invention is regulated
  • Figs. 2a-b illustrate schematic representation of the ROI in which the system of the present invention was trained
  • Figs. 3a-b illustrate the results of the analysis of the vegetation which was collected from Ramat Hagolan
  • Figs. 4a-f illustrate spectral analysis of vegetation
  • Figs. 5a-b illustrate special analysis of vegetation
  • Figs. 6a-b illustrate analysis of various materials
  • Figs. 7a-b illustrate matrix of the analysis of various materials
  • FIGs. 8a-b illustrate another embodiment of the present invention
  • Figs. 9a-b illustrate the MSSF and the CGSF of the present invention
  • Fig. 10 illustrates an embodiment of the algorithm of the present invention
  • Fig. 1 la-c illustrate various maps and sensing result provided by the system of the present invention
  • Fig. 12 illustrates spectral analysis performed by the system of the present invention
  • Figs. 13a-b illustrate the results of the system of the present invention
  • Fig. 14 illustrates the results of the system of the present invention
  • Fig. 15 illustrates the results of the system of the present invention
  • Fig. 16 illustrates the results of the system of the present invention
  • Fig. 17a-b illustrates the results of the system of the present invention
  • Fig. 18 schematically illustrates remote sensing of a minefield
  • Fig. 19 schematically illustrates spectra of radiation
  • Fig. 20 schematically illustrates simplified system and method
  • Fig. 21 schematically illustrates system and method incorporating auxiliary representation of the region of interest
  • Fig. 22 schematically illustrates system and method incorporating auxiliary identification of minefields
  • Fig. 23 schematically illustrates neural networks
  • Fig. 24 schematically illustrates neural networks.
  • the term 'sensing unit' refers hereinafter to any sensing mechanism which is adapted to sense an emitted electromagnetic radiation of an object.
  • the sensing unit may be for example any sensor which is able to detect electromagnetic radiation in the IR spectrum.
  • the sensing unit which is applied by the system of the present invention may be one of the following: ASTER sensor, AISA- Dual sensor, HyMap sensor, or any other hyperspectral imaging system in the range of 0.45 - 2.5 ⁇ .
  • the term 'ROT refers hereinafter to a region of interest.
  • the ROI may be any region with predetermined physical dimensions which may by sensed by the sensing unit of the present invention.
  • the term 'analyzing unit' refers hereinafter to a computation system which is adapted to interact with the sensing unit of the present invention for performing analysis of the data received from said sensing unit, and for providing a computational data to a user.
  • the term 'exposed chemical' refers hereinafter to a chemical substance located on a surface and exposed to view, rather than buried underground or hidden that is characterized by typical remotely visible property, for example in its color or in its spectrum of transmission, absorption or emission.
  • the term 'exposed chemical' may relate, for example and in a non-limiting manner to flavenoids, flavonols, anthocyanidins, and/or anthocyanins, wherein flavenoid is any plant secondary metabolites, defined according to the IUPAC nomenclature as (i) flavonoids, especially wherein the metabolite is derived from the 2-phenylchromone (2-phenyl-l,4-benzopyrone) structure; (iz) isoflavonoids, wherein the metabolite is derived from the 3-phenylchromone (3-phenyl-l,4-benzopyrone) structure; and (Hi) neofiavonoids, wherein the metabolite is derived from
  • the term may refer to any of the flavonoid aglycones, flavonoid O- glycosides, flavonoid C-glycosides, flavonoids with hydroxyland/or methoxy substitutions, C- methylflavonoids, methylenedioxyflavonoids, chalcones, aurones, dihydrochalcones, flavanones, dihydroflavanols, anthoclors, proanthocyanidins, condensed proanthocyanidins, leucoanthocyanidins, flavan-3,4-ols, flavan-3-ols, glycosylflavonoids, biflavonoids, triflavonoids, isoflavoneoids, isoflavones, isoflavanones, rotenonoids, pterocarpans, isoflavans, quinone derivatives, 3-aryl-4- hydroxycoumarins, 3-arylcoumarin, isoflav
  • Flavonols are any flavonoids possessing the 3- hydroxy-2-phenyl-4H-l-benzopyran-4-one backbone as defined by IUPAC. Their diversity stems from the different positions of the phenolic -OH groups, exemplified in a non-limiting manner by quercetin (3,5,7,3',4'-pentahydroxy-2-phenyl-4H-l-benzopyran-4-one), kaempferol (3,5,7,4'-tetrahydroxy-2- phenyl-4H-l-benzopyran-4-one) and myricetin (3,5,7,3',4',5'-hexahydroxy-2-phenyl-4H-l-benzopyran- 4-one).
  • Anthocyanidins are any flavenoid possessing an oxygen-containing heterocycle pyran fused to a benzene ring wherein the pyran ring is connected to a phenyl group at the 2-position, which can carry different substituents.
  • Anthocyanins are anthocyanidins possessing a sugar moiety.
  • the term 'spectral parameters' refers hereinafter to spectral features of the measured signals (e.g., images) received from the sensing unit of the present invention.
  • the spectral parameters may be used for the classification of said measured signals.
  • the spectral parameters may be associated with predetermined chemical characteristics of the exposed chemicals which are located in or absent from the ROI.
  • 'vegetation index' refers hereinafter to a single number that quantifies vegetation biomass and/or plant vigor for each pixel in a remote sensing image.
  • the index is computed using several spectral bands that are sensitive to plant biomass and vigor.
  • Vegetation Parameters are combinations of surface reflectance at two or more wavelengths designed to highlight a particular property of vegetation. They are derived using the reflectance properties of vegetation described in Plant Foliage. Each of the Vis is designed to accentuate a particular vegetation property.
  • 'SR', 'NDVI', 'PVF, 'SAVT, W, 'RDVF, *MSR', 'WDVF, 'MNLF, NDVI*SR, SAVI*SR, and 'TSAVF refer hereinafter to two-band vegetation parameters as described in tables 1 and 2 presented below:
  • the term 'green-ratio' refers hereinafter to crop vigour, vegetation amount or biomass, resulting from inputs, environmental, physical and cultural factors affecting crops.
  • the term 'SAM algorithm in SWIR2' refers hereinafter to an algorithm that determines the spectral similarity between two spectra by calculating the angle between the spectra, treating them as vectors in a space with dimensionality equal to the number of bands29,30. SAM compares the angle between the spectrum vectors of a known class with each pixel vector of an unknown class in n-dimensional space. SAM was performed using the selected endmembers of each species as different classes.
  • the term 'PPI' refers hereinafter to a widely used parameter for hyperspectral image analysis for endmember extraction due to its publicity and availability in the Environment for Visualizing Images (ENVI) software (Chang, 2006).
  • ENVI Environment for Visualizing Images
  • the PPI was implemented in the present invention to find end- members for the image scene, using the same set of randomly generated initial skewers.
  • MSSD minefield spectral signature dataset
  • the term may be specific for a set of spectral parameters which are implemented for the classification.
  • MSSF minefield spectral signature functions
  • control group signature function refers hereinafter to the function (or a plurality of functions) of the spectral signature of non-minefields images (control group) from which the MSSD is retrieved.
  • 'minefield' refers in the present invention to any geographical region comprising mines in its soil, while a mine comprise explosive materials (e.g., TNT) at least partially located in a container.
  • explosive materials e.g., TNT
  • the term 'propagation process' refers in the present invention to a chemical, geochemical, physical or biological process that propagates information concerning the existence of one material or activity at one location by causing the existence of another material or chemical at another location.
  • biological agent' refers in the present invention to a life-form capable of exercising propagation processes and producing exposed chemicals.
  • the term "biological agent' may be related inter alia to any plant, plant organ or tissue including without limitation, fruits, seeds, embryos, meristematic regions, callus tissue, flowers, leaves, roots, shoots, gametophytes, sporophytes pollen, and microspores, leaves etc; and to microorganisms, such as bacteria, algae, fungi etc.
  • the term "biological agent' may be related inter alia to microorganisms, such as bacteria, e.gi, pigment producing bacteria, cyanobacteria or photosynthetic bacteria, fungi, alga etc.
  • the microorganism may comprise of photosynthetic pigments, e.g., bactenochlorophyll; those bacteria are selected in a non-limiting manner from Purple bacteria; Green sulfur bacteria, Chloroflexi; Heliobacteria etc.
  • the method for minefield identification according to a most general embodiment of the present invention schematically characterized by a step of remote sensing a region of interest and the formation of an image, which is composed of picture elements, each element comprising a description of radiation intensity as a function of wavelength, a step of analyzing these descriptions of radiation to identify a property of these that may jserve to distinguish between minefield locations and locations relatively free of mines, and a step of analyzing an image and identifying minefield location therein.
  • This method can find such a property because of the activity of biological agents in the soil, which propagation processes produce detectable exposed chemicals in a manner dependant on the existence of mines in the soil, or more generally the activity of any other propagation process.
  • the present invention discloses a system and a method for identification and location of a minefield in a geographical region of interest (ROI).
  • the system comprises the following components:
  • a sensing unit for receiving the electromagnetic radiation of the ROI and generating at least one image of the electromagnetic radiation of the ROI.
  • the ROI is characterized by a specific geographical location.
  • an analyzing unit adapted to classify the probability of the at least one image to be a minefield by using one or more spectral parameters associated with at least one exposed chemical of the ROI.
  • the spectral parameters are selected from the group consisting of: SR, NDVI, PVI, SAVI, NLI, RDVI, MSR, WDVI, MNLI, NDVI*SR, SAVI*SR, TSAVI, green-ratio, Red-NIR parameters, NIR-SWIR1 slope, SAM algorithm in SWIR2, PPI, Nitrogen spectral index, and any combination thereof.
  • the object of the present invention is to develop a system and a method for minefields mapping using remote sensing technique.
  • the object of the present invention is to improve the speed, efficiency, and safety of Level 1 and 2 surveys (traditional method for mine field tracking and exposure) by analyzing high spectral and spatial imaging systems.
  • the present invention is adapted to perform the detection of minefields by spectral identification techniques and characterization of vegetation in the ROI.
  • the present invention is adapted to perform the detection of minefields by spectral identification techniques and characterization of soil in the ROI.
  • a system 100 for detection of minefields is employed.
  • a plant 200 is shown as an example of a biological agent, which may be influenced by a propagation process (e.g., a geochemical process).
  • the roots of plant 200 are located in the soil, possibly, in the vicinity of a mine.
  • Leakage from mine 500 changes the chemistry of the soil, and chemical compounds from the mine may find their way into the plant.
  • the mine may change a chemical process in plant 200, and thus cause a detectable change in the plant.
  • the color of flowers or leaves may change because of the near presents of a mine.
  • a sensing unit 400 is adapted to sense the electromagnetic radiation reflected from plant 200 and to generate at least one image of the electromagnetic radiation of the plant.
  • plant 200 is located in a specific ROI, so sensing unit 400 may receive the electromagnetic radiation of the ROI and generate at least one image of the electromagnetic radiation of the ROI.
  • the ROI is also characterized by a specific geographical location in which mine 500 and plant 200 are located.
  • Fig. la also illustrates an analyzing unit 300 which is adapted to classify the probability of the at least one image to be a minefield by using one or more spectral parameters associated with at least one exposed chemical of the ROI.
  • the exposed chemical are the chemicals of plant 20 which may be influenced to the leakage of different materials from mine 500.
  • analyzing unit 300 is adapted for performing the following procedues:
  • sensing unit 400 may comprise a sensor by which the images are acquired.
  • the sensor may be one of the following: ASTER sensor, AISA-Dual sensor, HyMap sensor.
  • sensing unit 400 may be connectable and transportable by a transportation object selected from the group consisting of: aircraft, satellite, ground vehicle, and any combination thereof.
  • sensing unit 400 further comprises at least one of: a satellite, an airborne camera, a camera carries by a land vehicle, and any combination thereof.
  • the exposed chemicals which are detectable by the system of the present invention may be produced by a propagation process depending on soil acidity.
  • the propagation process may also be depending on the concentration of molecules in the soil that are rich in either nitrogen or iron.
  • the exposed chemicals may produced by a geochemical propagation process.
  • the ROI may comprise biological agents (e.g., plants), and the spectral parameters are adapted to indicate at least one exposed chemical produced by the biological agents.
  • the biological organisms may further comprise at least one of: herbivores feeding on said plants and parasites on said plants.
  • the biological agents may comprise microorganisms (e.g., bacteria).
  • the analyzing unit is further adapted for identifying different types of soil and vegetation existing in the ROI; and identifying minefields per each identified type.
  • the new procedure based on airborne platform and the optimized set of hyper-, multi-, and digital high resolution sensors including Digital Terrain Model (DTM) allow a repetitive and fast survey for area reduction, increase the flexibility and adaptation to several scenarios, and above to all reduce the risk.
  • An enhanced GIS resulting in a 4-Level information and decision system based on satellite and airborne data, areal photography, ground investigation, GPS measurements and Level- 1 survey gathered information, will be developed.
  • the Expert System (ExS) for demining to be developed based on the existing GIS but it will also provide additional information from external sources and the results of the suggested system will be presented as a complete project.
  • the ExS consists in several components as illustrated in Fig. lb. The components are organized around the operational procedure that considered the following phases:
  • Information gathering phase - collection and storage of data in GIS (a) available information about the suspectred mined area from the general survey data, (b) available maps, (c) general area evaluation (terrain, vegetated area and temporal change detection).
  • Examination and planning phase will determine the site conditions (spectral ground truth data and spectral models), possible extent of the minefield and develop a plan for the airborne survey. Several aspects will be considered: the environment (weather, vegetation etc.) and mined area (evidence of mines, possible infiltration routes, or available minefield sketch in case of planned and maintained field).
  • Minefield survey phase The hyperspectral sensors (HyMap and AISA-Dual) will acquire images by the defined search pattern using the Google Earth interface. Repetitive pattern flights as well as different flight heights will be planed for spectral and multiscale image data analysis. All the acquired images and the corresponding metadata (flight path, Flight height, sensor parameters, recorded image position) will be then stored into the GIS for further processing.
  • the system of the present invention has to be trained.
  • a minefield spectral signature dataset (MSSD) is generated.
  • MSSD minefield spectral signature function
  • CGSF control group signature function
  • the MSSF and the CGSF are empiric and/or theoretic functions which can be created by performing different experiments in minefield and non-minefield locations.
  • the MSSD may be generated from experimental data received from a minefield (e.g., the Golan Heights in Israel), and a location which is not a minefield. Moreover, a laboratorial analysis of vegetation and/or soil from may be performed to evaluate the retrieval reflectance accuracy of the measures signals. The laboratorial analysis of the targets is helpful for the evaluation of the accuracy of retrieval reflectance of airborne AISA-Dual sensor and space-borne ASTER Level 2 data.
  • the Golan Heights are located between the borders of Iran, Riverside, Jordan, and Georgia. For that reason the Golan Heights have been a transit zone between these countries throughout history.
  • FIG. 2a presents an experimental area of primary airborne hyperspectral campaign with AISA-Dual sensor over selected minefield in Golan Heights. According to this figure the minefield is marked by a green circle.
  • the ground team was divided into two subdivisions: sampler and measurements.
  • the sampler entre to the minefield area and collect vegetated samples around the marked mines.
  • Each mine represented with three to four samples of green upper leaves.
  • This minefield defined by a standard mine laying method, we collect five mines samples for each of the four rows.
  • Isotopes are defined as atoms of the one element that differ in the number of neutrons present in their nuclei, i.e.have different mass numbers. All but 12 elements exist as mixtures of isotopes. Each element has a dominant light isotope (e.g. 12C (carbon), 14N (nitrogen), 160 (oxygen), 32S (sulphur), and 1H (hydrogen)), and one or two heavy isotopes (e.g. 13C, 15N, 170, 180, 33S, 34S, and 2H) with a natural abundance of a few percent or less.
  • a dominant light isotope e.g. 12C (carbon), 14N (nitrogen), 160 (oxygen), 32S (sulphur), and 1H (hydrogen)
  • one or two heavy isotopes e.g. 13C, 15N, 170, 180, 33S, 34S, and 2H
  • Table 3 displays relative abundances of naturally occurring isotopes of elements. Natural variations occur in the isotopic composition of lighter elements. These variations are due to fractionation effects, resulting in the creation of specific isotope ratio values that are characteristic of the origin, purity, and manufacturing processes of the products and their constituents.
  • the isotopic composition of plants and their constituents is based on a number of processes and environmental conditions. The significant variations were observed in the nitrogen isotope ratio values within the samples. This variation was suggested as possibly being indicative of different sources of nitrogen being used during the plant growth, i.e. TNT concentration within minefield soils.
  • the nitrogen isotope ratios of TNT obtained from nitro groups could be differentiated from the nitrogen isotope ratios from amino groups.
  • Fig 3b illustrates the results of chemical isotopic analysis for N/N15 evaluation. According to this figure, the variation between plants within and outside minefield becomes apparent and significant. An average value of measured nitrogen ratio within minefields is 2.5, while control grout characterized with value of 1.2. This analysis conform the hypothesis of ExS project that plants within minefields are chemically contaminant by the residuals of TNT.
  • Spectral analysis The suggested spectral analysis focused on separation between vegetation within minefield and the control group. Several preparations have been performed prior spectral analysis. Spectral calibration process standardized and normalized by measurements of an internal standard. This process enable the isolation of noisy wavelengths (from signal) and a generation of noisy-less smooth data set for further analysis.
  • the reflectance spectra of internal standard and measured reflectance spectra of the plants were normalized to the continuum level by interactively choosing continuum points, linearly interpolating between ⁇ hem, and then dividing the final spectrum by the continuum. Equivalent widths (EW) were measured by choosing the linear boundaries interactively. The calculated ratio between continuums removed spectra of internal standard provides us a normalization factor. For each mine detected on field three samples of plants were collected and each plant was measured spectrally with three replications.
  • Figs. 4a-b illustrates the average spectra of minefield and control group plants. As it seen in Figs. 4a-b the minefield samples demonstrates higher internal similarity than control group that shows high variation. Although, the differences between average spectra were not significant, we decided to investigate this phenomenon.
  • the minefield spectra were divided into four groups as a number of selected mine rows in the field, with five averaged spectra in each row as a number of sampled mines.
  • Figs. 4c-f present the spectra of four groups accompanied by internal group average spectrum (red curve).
  • Figs. 4c-f illustrate the spectra of four mine rows demonstrate the inter-connection between spectrum and the location related to the center of the minefield. As the row closer to the center the spectra gathered closer to the group average and minimize spectral variation.
  • Spatial analysis The spatial analysis executed on spectral in situ measurements based on simulated image. This image converts each spectrum into single pixel with spatial/spectral profiles.
  • Figure 5a-b presents created image, while columns replicate mine rows (four rows) and across dimension (image rows) imitate mines in the same mine row. In case of control group pixels order have no spatial meaning. Color index of both stages of vegetation (healthy green and dry) are provided, when color of green plants varied between light and dark green establish plants pigment in VIS region, as the pigment higher the color is brighter that interpretative as healthy vegetation and vice versa for less healthy vegetation that interpretive by dark pigment.
  • Figs. 5a-b illustrate simulated images of spectral measured plants in two stages, green and healthy vegetation measured in situ and after drying process under laboratory conditions.
  • Remote sensing has proven a powerful "tool” for assessing the identity, characteristics, and growth potential of most kinds of vegetative matter. Vegetation behavior depends on the nature of the vegetation itself, its interactions with solar radiation and other climate factors, and the availability of chemical nutrients and water within the host medium. Because many remote sensing devices operate in the green, red, and near infrared regions of the electromagnetic spectrum, they can discriminate radiation absorption and reflectance of vegetation.
  • a photosystem has two closely linked components, an antenna containing light-absorbing pigments and a reaction center comprising a complex of proteins and two chlorophyll a molecules. Each antenna contains one or more light-harvesting complexes (LHCs). The energy of the light captured by LHCs is funneled to the two chlorophylls in the reaction center, where the primary events of photosynthesis occur.
  • LHCs light-harvesting complexes
  • chlorophyll a (or any other molecule) absorbs VIS light, the absorbed light energy raises the chlorophyll a to a higher energy state, termed an excited state. This differs from the ground state largely in the distribution of electrons around the C and N atoms of the porphyrin ring. Excited states are unstable, and will return to the ground state by one of several competing processes.
  • a vegetation index is a single number that quantifies vegetation biomass and/or plant vigor for each pixel in a remote sensing image.
  • the index is computed using several spectral bands that are sensitive to plant biomass and vigor.
  • Vegetation Parameters are combinations of surface reflectance at two or more wavelengths designed to highlight a particular property of vegetation. They are derived using the reflectance properties of vegetation described in Plant Foliage. Each of the Vis is designed to accentuate a particular vegetation property. More than 150 Vis have been published in scientific literature, but only a small subset have substantial biophysical basis or have been systematically tested. The parameters are grouped into categories that calculate similar properties :
  • Narrowband Greenness - combination of reflectance measurements sensitive to the combined effects of foliage chlorophyll concentration, canopy leaf area, foliage clumping, and canopy architecture;
  • Light use efficiency - provide a measure of the efficiency with which vegetation is able to use incident light for photosynthesis
  • Canopy Nitrogen designed to provide a measure of nitrogen concentration of remotely sensed foliage
  • Dry or Senescent Carbon designed to provide an estimate of the amount of carbon in dry states of lignin and cellulose
  • Leaf Pigments - provide a measure of stress-related pigments present in vegetation
  • Canopy Water Content provide, a measure of the amount of water contained in the foliage canopy.
  • the vegetation indexed applied in the present invention are presented in Tables 1 and 2 above.
  • the analyzing unit adapted to classify the probability of said ROI to be a minefield according to a classification technique selected from the group consisting of: K-mean, ISODATA, Spectral Angle Mapper, Artificial Neural Network, Pattern recognition, genetic algorithm, and any combination thereof.
  • the new spatial dimension added to spectral measurements provides an opportunity to use well-known classification algorithms for spectral separation between minefield and control group vegetation. This procedure based on supervised and unsupervised techniques.
  • Classification process define any individual pixel or spatially grouped sets of pixels as representing some feature, class, or material is characterized by a range of spectral signatures monitored by the wavelengths of provided remote sensor.
  • the spectra are considered to be clustered sets of data in 2-, 3-, and higher dimensional plotting space. These are analyzed statistically to determine their degree of uniqueness in this spectral response space and some mathematical functions are chosen to discriminate the resulting clusters.
  • Two methods of classification are commonly used: Unsupervised and Supervised. In unsupervised classification any individual pixel is compared to each discrete cluster to see which one it is closest to. A map of all pixels in the image, classified as to which cluster each pixel is most likely to belong, is produced. This then must be interpreted by the user as to what the color patterns may mean in terms of classes, etc. that are actually present in the real world scene; this requires some knowledge of the scene's feature/class/material content from general experience or personal familiarity with the area imaged.
  • a supervised classification the interpreter knows beforehand what classes, etc. are present and where each is in one to perhaps many locations within the scene. These are located on the image, areas containing examples of the class are; circumscribed making them training sites, and the statistical analysis is performed on the multiband data for each such class. Instead of clusters then, one has class groupings with appropriate discriminant functions that distinguish each. All pixels in the image lying outside training sites are then compared with the class discriminants derived from the training sites, with each being assigned to the class it is closest to - this makes a map of established classes with a few pixels usually remaining unknown, which can be reasonably accurate but some classes present may not have been set up; or some pixels are misclassified.
  • the two most frequently used algorithms are the K-mean and the ISODATA clustering algorithm. Both of these algorithms are iterative procedures. In general, both of them assign first an arbitrary initial cluster vector. The second step classifies each pixel to the closest cluster. In the third step the new cluster mean vectors are calculated based on all the pixels in one cluster. The second and third steps are repeated until the "change" between the iteration is small.
  • the "change” can be defined in several different ways; either by measuring the distances the mean cluster vector has changed from iteration to another or by the percentage of pixels that have changed between iterations.
  • the ISODATA algorithm has some further refinements by splitting and merging of clusters (Jensen, 1996).
  • Clusters are merged if either the number of members (pixel) in a cluster is less than a certain threshold or if the centers of two clusters are closer than a certain threshold. Clusters are split into two different clusters if the cluster standard deviation exceeds a predefined value and the number of members (pixels) is twice the threshold for the minimum number of members.
  • the ISODATA algorithm is similar to the k-means algorithm with the distinct difference that the ISODATA algorithm allows for different number of clusters while the k-means assumes that the number of clusters is known a priori.
  • Supervised classification is much more accurate for mapping classes, but depends heavily on the cognition and skills of the image specialist.
  • the strategy is simple: the specialist must recognize conventional classes or meaningful classes in a scene from prior knowledge, such as personal experience with what's present in the scene, or more generally, the region it's located in, by experience with thematic maps, or by on-site visits.
  • Fig. 6a schematically illustrates Minimum a distance diagram for simplicity demonstrated only by two damnations.
  • the Minimum Distance routine presented in Fig. 6 involves fairly simple calculation procedures. The data points from different wavelengths are dots; the mean for each clustered data set are square. For point 1, an unknown, the shortest straight-line distance to the several means is to the class "heather”. Point 1, then, is assigned to this category.
  • a statistical Bayesian Probability Function is also calculated for the class data.
  • Fig. 6b illustrated a scatter plot that shows the outer envelope bounding each class in Maximum likelihood algorithm.
  • the class spectral distribution illustrated in Fig. 6b again for a two-dimensional case, gives rise to elliptical boundaries which define the equiprobability envelope for each class.
  • the Spectral Angle Mapper is an automated method for comparing image spectra to individual spectra or a spectral library (Boardman, unpublished data; CSES, 1992; Kruse et al ., 1993a). SAM assumes that the data have been reduced to apparent reflectance (true reflectance multiplied by some unknown gain factor controlled by topography and shadows). The algorithm determines the similarity between two spectra by calculating the "spectral angle" between them, treating them as vectors in a space with dimensionality equal to the number of bands. A simplified explanation of this can be given by considering a reference spectrum and an unknown spectrum from two-band data.
  • the two different materials will be represented in the 2-D scatter plot by a point for each given illumination, or as a line (vector) for all possible illuminations. Because it uses only the "direction" of the spectra, and not their “length,” the method is insensitive to the unknown gain factor, and all possible illuminations are treated equally. Poorly illuminated pixels will fall closer to the origin.
  • the "color” of a material is defined by the direction of its unit vector. Notice that the angle between the vectors is the same regardless of the length. The length of the vector relates only to how fully the pixel is illuminated
  • Partial Least Squares The PLS regression technique is especially useful in quite common case where the number of descriptors (independent variables) is comparable to or greater than the number of compounds (data points) and/or there exist other factors leading to correlations between variables. In this case the solution of classical least squares problem does not exist or is unstable and unreliable.
  • PLS approach leads to stable, correct and highly predictive models even for correlated descriptors [Martens 1989, H6skuldsson 1988, Eriksson 2001].
  • MFTA Molecular Field Topology Analysis
  • Partial Least Squares regression is based on linear transition from a large number of original descriptors to a new variable space based on small number of orthogonal factors (latent variables).
  • factors are mutually independent (orthogonal) linear combinations of original descriptors.
  • latent variables are chosen in such a way as to provide maximum correlation with dependent variable; thus, PLS model contains the smallest necessary number of factors [2]. With increasing number of factors, PLS model converges to ordinary multiple linear regression model (if one exists). In addition, PLS approach allows one to detect relationship between activity and descriptors even if key descriptors have little contribution to the first few principal components.
  • Model predictivity The approach to factor construction provides the description of available data using minimum number of adjustable parameters and, consequently, maximum precision and stability of regression model. However, inclusion of excessive factors in the model increases the accuracy of description but may decrease the predictivity as model starts to represent not just the true pattern of relation between descriptors and activity but also random noise and individual features of the training set. Because of this, during construction of the model its predictivity is monitored after including each successive factor by means of cross-validation procedure. In cross-validation approach, computation is run several times in such a way that certain subset of the training set is not used in the model construction. Then the activity is predicted for excluded compounds using such partial model. Each compound is excluded exactly once, and normalized total error of prediction for them serves as a measure of predictivity for the full model - cross-validation parameter Q2 that is used in PLS regression to select optimal number of PLS factors.
  • the first step consists in the elimination of the low-variable (almost constant) descriptors that are different from a constant only for a few (2-3) compounds in the training set.
  • descriptors cannot provide useful statistical information. In most cases, they simply help to artificially fit these particular compounds, thus decreasing the model predictivity.
  • the results of PLS model suggested possible overfitting that traditionally defined as training some flexible representation so that it memorizes the data but fails to predict well in the future. Overfitting occurs when the models describe the examples better and better but get worse and worse on other instances of the same phenomenon. It makes the whole learning process worthless. We observe overfitting by splitting a number of examples in two, a training set, and a test set and training the models on the training set.
  • ANN Artificial Neural Networks
  • An Artificial Neural Network is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information.
  • the key element of this paradigm is the novel structure of the information processing system. It is composed of a large number of highly interconnected processing elements (neurones) working in unison to solve specific problems.
  • ANNs like people, learn by example.
  • An ANN is configured for a specific application, such as pattern recognition or data classification, through a learning process. Learning in biological systems involves adjustments to the synaptic connections that exist between the neurons [Aleksander].
  • Pattern Recognition An important application of neural networks is pattern recognition. Pattern recognition can be implemented by using a feed-forward neural network that has been trained accordingly.
  • the network is trained to associate outputs with input patterns.
  • the network identifies the input pattern and tries to output the associated output pattern.
  • the power of neural networks comes to life when a pattern that has no output associated with it, is given as an input. In this case, the network gives the output that corresponds to a taught input pattern that is least different from the given pattern.
  • Feedback networks having signals travelling in both directions by introducing loops in the network.
  • Feedback networks are very powerful and can get extremely complicated.
  • Feedback networks are dynamic; their 'state' is changing continuously until they reach an equilibrium point. They remain at the equilibrium point until the input changes and a new equilibrium needs to be found.
  • Feedback architectures are also referred to as interactive or recurrent, although the latter term is often used to denote feedback connections in single-layer organizations.
  • the Learning Process An associative mapping in which the network learns to produce a particular pattern on the set of input units whenever another particular pattern is applied on the set of input units.
  • An auto-association determines an input pattern is associated with itself and the states of input and output units coincide. This is used to provide pattern complete, i.e. to produce a pattern whenever a portion of it or a distorted pattern is presented.
  • the network actually stores pairs of patterns building an association between two sets of patterns.
  • the weights are fixed a priori according to the problem to solve was determined.
  • Our model was supervised learning which incorporates an external teacher, so that each output unit is told what its desired response to input signals ought to be. During the learning process global information may be required. Paradigms of supervised learning include error-correction learning, reinforcement learning and stochastic learning. Important issue concerning supervised learning is the problem of error convergence, i.e. the minimization of error between the desired and computed unit values. The aim is to determine a set of weights which minimizes the error.
  • LMS least mean square
  • ANN The results of ANN perform with specific task, following the selection of how the units are connected to one another, and accurate setting of weights on the connections presented in Figure 20.
  • the connections determine whether it is possible for one unit to influence another.
  • the weights specify the strength of the influence.
  • a three-layer 1 network has performed a particular task by using the following procedure:
  • the results will be presented as a confusion matrix that is a visualization tool typically used in supervised learning methods. Each row of the matrix represents the instances in a predicted class, while each column represents the instances in an actual class.
  • a confusion matrix is that it is easy to see if the system is confusing two classes (i.e. commonly mislabeling one as another. When the data set is unbalanced the error rate of a classifier is not representative of the true performance of the classifier.
  • Fig. 7a-b illustrated the results of the suggested AN s models (left 3 neuron model, right 2 neuron model).
  • the ANNs models in this figure are summarized by a confusion matrix. It is clear that the 3 neuron model is more complicated model; additionally the presented results illustrate better performance for 2 neuron model.
  • the total success of 2 neuron model is 92.5% that classified 1 target (reflectance spectrum) from minefield as a control group plant, and 2 targets of control group interpolated as minefield vegetation.
  • the significant of 3 neuron model is 90% with 10% of errors and mistakes, while 2 targets of control group were classified into minefield class and 2 another targets of minefield were included in control class.
  • the results of both ANNs models are significant and outstanding.
  • GAC Genetic algorithm - Genetic algorithms
  • a population of abstract representations called chromosomes or the genotype of the genome
  • candidate solutions called individuals, creatures, or phenotypes
  • the evolution usually starts from a population of randomly generated individuals and happens in generations. In each generation, the fitness of every individual in the population is evaluated, multiple individuals are stochastically selected from the current population (based on their fitness), and modified (recombined and possibly randomly mutated) to form a new population. The new population is then used in the next iteration of the algorithm. Commonly, the algorithm terminates when either a maximum number of generations has been produced, or a satisfactory fitness level has been reached for the population.
  • the results of the suggested GA presented in Fig. 7c.
  • the blue curve implements calculated cases and red curve represents verification (measured cases).
  • the suggested GA model was trained with 10 selected cases of minefield and control group vegetation spectra, and,/ tested on others independent 10 cases.
  • the results shows two curves represented real measurements (red) and calculated labels (classification and separation) blue curve.
  • the labels illustrate classes, when 0 means control group and 1 means minefield plants. It is clearly seen that only in one case the label of calculated minefield plant does not match the validation curve of measured cases (red curve).
  • One mistake of 10 tested cases provides an outstanding accuracy of 90%.
  • the scheme of present invention contains following stages:
  • An airborne Hyperspectral campaigns were executed to estimate the probability for detection of buried mines using an AISA-Dual and HyMap hyperspectral imagery scanned from airborne platforms.
  • a Geographic Information System (GIS), integrated with hyperspectral imagery and classifications layers, and including an advanced applications, should allow foe fusion of measured image data, a priori information (ground truth) and geographic information to be used for planning of demining activities and as working support during operational process. Validation in controlled environment and real minefield will allow the suggested method to achieve effective results for Level 2 survey in a way which is acceptable for minefield exposure.
  • ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer
  • the ASTER instrument consists of three separate instrument subsystems. Each subsystem operates in a different spectral region, has its own telescope(s), and was built by a different Japanese company.
  • ASTER's three subsystems are: the Visible and Near Infrared (VNIR), the Shortwave Infrared (SWIR), and the Thermal Infrared (TIR).
  • VNIR Visible and Near Infrared
  • SWIR Shortwave Infrared
  • TIR Thermal Infrared
  • the VNIR subsystem operates in three spectral bands at visible and near-IR wavelengths, with a resolution of 15 m.
  • the backward-looking telescope provides a second view of the target area in Band 3 for stereo observations.
  • Thermal control of the CCD detectors is provided by a platform- provided cold plate.
  • Cross-track pointing to 24 degrees on either side of the track is accomplished by rotating the entire telescope assembly.
  • Band separation is through a combination of dichroic elements and interference filters that allow all three bands to view the same ground area simultaneously.
  • the data rate is 62 Mbps when all four bands are operating.
  • Two on-board halogen lamps are used for calibration of the nadir-looking detectors. This calibration source is always in the optical path.
  • the SWIR subsystem operates in six spectral bands in the near-IR region through a single, nadir-pointing telescope that provides 30 m resolution.
  • Cross-track pointing ( ⁇ 8.550) is accomplished by a pointing mirror. Because of the size of the detector/filter combination, the detectors must be widely spaced, causing a parallax error of about 0.5 pixels per 900 m of elevation. This error is correctable if elevation data, such as a DEM, are available.
  • Two on-board halogen lamps are used for calibration in a manner similar to that used for the VNIR subsystem, however, the pointing mirror must turn to see the calibration source. The maximum data rate is 23 Mbps.
  • Satellite image digital numbers cannot be assumed to represent actual surface conditions because of a variety of effects, such as variable atmospheric attenuation, illumination geometry, and sensor characteristics. Therefore, the quantitative utility of remotely sensed data is maximized by calibrating it to a surface reflectance factor.
  • the empirical line (EL) atmospheric correction technique is a common and effective way of correcting multispectral and hyperspectral data from raw DNs, or radiance, to reflectance factors (Smith and Milton 1999, Clark et al. 2002, Ben-Dor et al. 2004). It assumes that a linear relationship exists between image DNs and ground measured reflectance for surfaces with a range of contrasting albedo ( Figure 8a). This linear relationship is used to calculate gains and offsets that convert DNs to reflectance factors.
  • the two-target approach was a generally accepted protocol in early applications of the EL method.
  • the process is: (i) homogeneous areas are identified on existing imagery of the field site; (ii) the target areas are visited by a field crew with a spectrometer to collect a suite of representative spectra; (iii) field spectra are matched with corresponding image pixels; and (iv) the EL is calculated. Acquisition of spectra is best undertaken simultaneously with the overpass of the sensor.
  • calibration sites should be at an elevation similar to the areas of interest in the image. Examples include: playas, lava flows, rock outcrops, bare dirt and gravel areas, or dam faces.
  • Atmospheric correction of Level B performed with atmospheric model.
  • the input parameters for the radiative transfer code include solar and view zenith angle and surface reflectance. Other inputs describe atmospheric properties, including molecular and aerosol scattering optical depths, aerosol index of refraction, and aerosol size distribution.
  • the code handles ozone absorption but strong absorption by water vapor must be treated separately.
  • the LUT (look-up tables) is based on the following parameters: solar zenith angle, satellite view angle, relative azimuth angle between the satellite and sun, molecular scattering optical depth, aerosol scattering optical depth, aerosol single scatter albedo, aerosol size distribution parameter, and surface reflectance.
  • the size distributions that are used are based either on a Junge size distribution (or power law) or are based on the set of aerosol types used to develop the atmospheric correction for MISR.
  • the Junge-based distributions are useful because they allow the size distribution to be described with one parameter. This simplifies the look-up procedures and reduces the size of and the time to generate the LUT. If, for instance, a log-normal distribution were used, two parameters would be required to generate and use the table.
  • the actual atmospheric correction is the process of retrieving the surface radiance and surface reflectance from the satellite radiances.
  • the scattering and absorbing properties of the atmosphere are determined using techniques described in the next section. Columnar amounts of absorbing gases are used to compute the sun-to-surface-to-satellite gaseous transmittance.
  • the satellite radiances are divided by these transmittances to determine satellite radiance for an atmosphere which does not contain gaseous absorption.
  • the scattering optical depths, single scatter albedo, and size distribution are used along with the solar and view angles to determine the appropriate portion of the LUT to use.
  • the values of the input parameters to the atmospheric correction will not match those used to generate the LUT.
  • a combination of interpolation and nearest-neighbor approaches will be used.
  • the advantage to interpolation schemes is that the size of the table may be reduced. However, this is at the cost of increasing the complexity of the LUT scheme. If a nearest-neighbor approach is used, the resolution of the table parameters must be selected so that the use of the table is not a primary source of error.
  • the LUT access is simplified, but the size of the table, and time needed to generate it, could become prohibitively large.
  • Version 1 we have adopted a nearest-case approach. This allows the code developers to test computational requirements of the code. It should be emphasized that the Version 2 code will use a combination of nearest-neighbor and interpolation. Once the appropriate surface radiances, reflectances, and TOA radiances have been determined from the table, piecewise-linear fits are calculated between the TOA radiance and the surface reflectance. Similar fits are also found between the TOA radiance and the surface radiance. The ASTER scene radiance is then used with the proper . linear fit to determine the surface radiance and reflectance. In Fig. 8b illustrated ASTER bands on top of atmospheric transmittance curve.
  • Fig. 9a illustrates two spectra of minefield and control group vegetation.
  • the minefield vegetation spectrum is colored in green, and it represents the MSSF function.
  • the control group vegetation is colored in blue, and it represents the CGSF.
  • Fig. 9b illustrates another representation of minefield spectra Vs. wavelength.
  • the MSSF has (x,y) values of about (0.5, 0.88), (0.635, 0.075), or any combination thereof.
  • the spectral parameters may be vegetation parameters and/or soil parameters.
  • the analyzing unit is further adapted to detect an existence of vegetation or soil in the ROI, and to perform the classification accordingly.
  • the first difference is inNIR region (band 3) when green and healthy vegetation presents low albedo in NIR region that disturbed results of spectral parameters e.g. NDVI, EVI and Red Edge, mentioned in table 1.
  • the third difference appears in SWIR2 region emphasize several spectral variation of relatively high reflectance of band 7, spectral features with low accuracy, and variation of local spectral slopes. Following parameters were developed: 1. Green Ratio index to detected healthy green vegetation, 2. Red-NIR parameters (NDVI and EVI) to determine albedo differences of band 3, 3.
  • NIR-SWIR1 slope to complete the disparity between band 3 and 4, 4.
  • SAM algorithm in SWIR2 region to emphasize an exclusivity of minefield vegetation.
  • the starting point of this algorithm is fully completed image preprocessing.
  • ASTER images have to be calibrated and corrected with radiometric coefficients calculated by EL method.
  • Atmospheric correction of the data executed with model-based and EL methods.
  • Geometric correction of the images based on Level B rectification method and evaluated by mid to high accuracy.
  • the spectral parameters may be normalized with chlorophyll absorption in 680 nm, and/or with liquid water absorptions in 940 nm and 1140nm.
  • the spectral analysis of ASTER data follows the Golan Heights algorithm by starting with image preprocessing stage that included data calibration and correction, geo-rectification and cloud masking (spectral data is not available under clouds and shadowed areas). After isolation of vegetated areas, the suggested algorithm for suspected mines areas completed by using four spectral parameters.
  • Figure 12 presents spectral similarity of suspected minefields areas in Nzeto against Golan Heights validated minefield.
  • the second areas of interest allow ground validation of algorithm detection.
  • the suspected minefields areas are located in Malanje and Dundo.
  • First ASTER images of Malanje area were collected and preprocessed.
  • the suggested algorithm provided thematic map of selected polygon and classified the suspected mines areas in following significant rates and detection accuracies that defined by similarity with Golan Heights spectral features: red class is the highest similarity (up to 90%), green class is mid similarity (about 60%), and blue class is relatively low similarity that is estimated under 50%.
  • Figure 13a shows thematic map of ASTER image with three detection classes of suspected mines accompanying with results of Norway mine detection group. The discovered mines and explosives were aged as more then ten years old and detected by conventional method.
  • the data of Norway group validates and approves spectral results of the suggested algorithm.
  • Accordign to Fig. 13 ASTER image classification (red heights accuracy, green mid and blue is relatively low accuracy) combined with planned Aldeia Nova complex area in Malanje and photos of mines and explosives that were discovered by conventional field method of Norway group.
  • the analyzing unit of the present invention is adapted to construct a map with the ROIs and their probability to be a minefield according to a predetermined scale.
  • the scale may be presented in different colors (e.g., red dots, blue dots, etc.).
  • the generation of the map may be performed by superposition of the probabilities' map and the geographical map of the inspected region. Additional area of interest of approximately 50 sq km was located in Dundo.
  • the results of the suggested spectral algorithm were clear of mines area as none of examined pixels were classified in any of detection classes.
  • the ground validation has confirmed the spectral results as no mined or any explosive has been found during conventional field method.
  • the suggested algorithm was accepted on ASTER images of Nagorno Karabakh area that located between Armenia and Azerbaijan. The results of this experiment are illustrated in Fig. 13b. The suspected minefields areas are located in Agdam and Tartar that were the center of national conflict more then 15 years ago.
  • HyMap - Hvperspectral Airborne Sensor - HyMap is a state-of-the-art aircraft-mounted commercial hyperspectral sensor developed by Integrated Spectronics, Sydney, Australia, and operated by HyVista Corporation. HyMap provides unprecedented spatial, spectral and radiometric excellence (Cocks, 1998).
  • the system is a whiskbroom scanner utilizing diffraction gratings and four 32-element detector arrays (1 Si, 3 liquidnitrogen- cooled InSb) to provide 125 spectral channels covering the 0.45 - 2.5 um range over a 512-pixel swath.
  • Table 4 summarizes the spectral characteristics for the four spectrometers. The Y-direction of the imagery is provided by the aircraft motion. During the 2009 Angola campaign, data were collected with spatial resolutions of 3 meters.
  • the HyMap series of sensors has attained a high level of performance in spectral and radiometric calibration accuracy, signal to noise ratio, geometric characteristics, and operational stability.
  • the HyMap sensor is spectrally calibrated in the laboratory by programming it to view a 150 mm diameter beam from a collimator directed into the sensor.
  • a monochromator and QI light source are used to illuminate (via an optical fiber coupling) the field stop of the collimator, the monochromator is scanned in wavelength, and the sensor records the signal levels at each wavelength.
  • Wavelength positions of emission lines from a Ne lamp or the absorption band positions present in a plastic with known absorption features are measured and verified with a Nicolet FTIR.
  • Spectral calibration procedures are used to determine the band center wavelengths and bandwidths. Overall spectral calibration accuracy is estimated as better than 0.5 nm.
  • Radiometric Calibration - The HyMap sensor is radiometrically calibrated by having the sensor directly view a 250mm square pad of Spectralon (SRT-99-100 from Labsphere) which is illuminated by a 1000 W QI lamp (Optronics Laboratories FEL-C). The lamp is about 650 mm from the pad and illuminates it at normal incidence. The sensor views the pad at an angle of 45 degrees.
  • the calibration data for the Spectralon pad and FEL lamp are supplied by the manufactures and are traceable to NIST standards. At this time, the largest source of radiometric calibration error (particularly in the SWIR region) is the uncertainty in the lamp calibration information supplied by the manufacturer.
  • SNR Signal-to-Noise Ratio
  • HyMap sensors have consistently delivered outstanding SNR performance. SNR achieved in lab by the sensor while imaging a 50% reflectance target and solar zenith angle of 30 degrees. Comparative analysis of dark images (with the sensor's shutter closed) recorded on the ground using laboratory power and during flight has not detected any significant levels of aircraft vibrational or electronic noise. Typically, the dark current noise varies by only about 0.1 DN between laboratory and airborne measurements, thus ground-based measurements of SNR are a very good indicator of the sensor's in-flight performance.
  • a Zeiss-Jena SM2000 stabilized platform is used to provide 1st level image geometry information.
  • the system is hydraulically actuated and provides +/- 5 degrees of pitch and roll correction.
  • the yaw can be offset by +/- 20 degrees with +/- 8 degrees of stabilization.
  • the yaw offset or drift is currently set manually by the operator.
  • the platform provides a residual error in nadir pointing of less than 1 degree and reduces aircraft motion effects by a factor ranging from 10:1 to 30:1.
  • a state-of-the-art GPS/INS "C-MIGITS II" system developed by Boeing was installed on HyMap starting with the 1999 flight season.
  • This device provides realtime delivery of position and attitude using combined Digital Quartz Inertial Measurement Units and a MicroTracker Global Positioning System receiver.
  • the CMIGITS II device was bolted directly to HyMap, measuring its position and three axes of rotation at rate of 10 Hz.
  • the position information from the GPS system and the 3- axis attitude data from the INS accelerometers are combined in a tightly coupled Kalman filter approach to give real-time information on x, y and z position values as well as roll, pitch and true heading.
  • the stated accuracy is 4.5 meters and 1.0 milliradian when operated with differential GPS.
  • Data Processing - Preprocessing consisted of making duplicates of aircraft tapes, reading the data from tape, performing the radiance calibration, correction to apparent reflectance, performing the "EFFORT” correction, calculating the geometric correction and producing a geometrically-corrected quick-look image, and writing the standard product Radiance and Reflectance CDs.
  • Geometric Correction - Routine delivery of geometrically corrected hyperspectral data is now possible beginning with the HyMap data acquisition (Boardman, 1999).
  • the HyMap geostabilized platform and high quality differential GPS augmented with a Boeing CMIGITS GPS/INS and post-processing are used to establish precise geometrically corrected true color image (quick-look), geometric correction factors, and software to convert the data to map coordinates were provided to the participants with each HyMap dataset. This allows users to analyze their data and then to convert analysis results to map- based products.
  • HyMap data was based on several spectral parameters that emphasizes and evaluates the uniqueness of minefield vegetation spectrum. The classification stages based on principal differences between spectra within minefields and control groups, as it has been done for ASTER images. In case of hyperspectral dataset the differences can be identified more detailed.
  • ExS Expert System
  • the main purpose of the suggested method is development of Expert System (ExS) that will provide a complete project that provided information, cross merging of maps, image maps and thematic information of GIS multi-layers project and precise information for decision support.
  • the spectral classification follows spectral models that were developed previously. The 12 wavelengths were selected by the models as a most significant and "interest" spectral regions.
  • the first stage of algorithm focused ⁇ spectral separation of relevant for this method grass vegetation and spatial masking of irrelevant vegetation e.g. trees and bushes. This has been done by two co-stages: first the NDVI (see table 1) separate vegetated areas, second PPI classify between relevant and irrelevant vegetation.
  • the Pure Pixels Index has been widely used in hyperspectral image analysis for endmember extraction due to its publicity and availability in the Environment for Visualizing Images (ENVI) software (Chang, 2006).
  • the PPI was implemented to find end-members for the image scene, using the same set of randomly generated initial skewers.
  • the spectral parameters cauterized and normalized to establish generic algorithm The first cluster combined VIS-NIR regions by using NDVI and VIS pigments parameters (see table 1) normalized with chlorophyll absorption in 680nm.
  • the second cluster included NIR-SWIR regions with Red Edge and Nitrogen spectral parameters (see table 1) normalized by liquid water absorptions in 940nm and 1140nm.
  • Figure 32 shows thematic map of HyMap image with two detection categories of suspected mines accompanying with results of Norway mine detection group.
  • the discovered mines and explosives were aged as more then ten years old and detected by conventional method.
  • the data of Norway group validates and approves spectral results of the suggested algorithm.
  • the detection included: mines, and unexploded explosive ordnance (UXO). All munitions containing explosives, nuclear fission or fusion materials, and biological and chemical agents. This includes bombs and warheads; guided and ballistic missiles; artillery, mortar, rocket, and small arms ammunition; all mines, torpedoes, and depth charges; demolition charges; pyrotechnics; clusters and dispensers; cartridge and propellant actuated devices;
  • FIG. 15 illustrates thematic map HyMap image of suspected mines accompanying with results of Norway group.
  • the total area of suspected minefield that overlaps with planned project of "Aldeia Nova" marked as a blue polygon.
  • most of the area that marked with a blue polygon free of mines and stays unclassified and undetected with both methods.
  • the matching between field detected materials and spectrally classified polygons is clear.
  • Fig. 16 presents results of spectral classification and field detection, while the distances between detected explosives and classified polygons is minimal and control area that is free of mines stays clear in both methods. The accuracy of spectral classification is very high and the significant of the suggested algorithms conformed and validated by field detections.
  • AISA-Dual - Hyperspectral Airborne Sensor - AISA-Dual is a compact sensor, which requires minimum maintenance, and is easy to install in small aircrafts.
  • the sensor assembly will only require a single hole in the aircraft, and fits on a standard stabilized camera platform.
  • the Hawk sensor can be aligned with respect the Eagle to make the both sensors to look at the same swath on the ground.
  • the Hawk fore lens is adjustable to match the SWIR ground pixel size to that in VNIR.
  • the new procedure based on airborne and space-borne platforms and the optimized set of hyper-, multi-, and digital high resolution sensors including Digital Terrain Model (DTM) allow a repetitive and fast survey for area reduction, increase the flexibility and adaptation to several scenarios, and above to all reduce the risk.
  • DTM Digital Terrain Model
  • An enhanced GIS resulting in a 4-Level information and decision system based on satellite and airborne data, areal photography, ground investigation, GPS measurements and Level- 1 survey gathered information, was developed.
  • the Expert System (ExS) developed based on the existing GIS program provide additional information from external sources and the results of the suggested system presented as a complete project. The multiplicity of experimental areas all over the glob clearly demonstrates the robustness of developed methodology.
  • the ExS method has a full potential to provide a useful addition to existing procedures and technology to solve the mine problem as currently exists in many mine infected countries.
  • the use of airborne sensors allows flexible approach in flight altitude and patterns.
  • the presented experiments of different campaigns, sensors and locations are und: ertaken to proof and optimize the method to true suspected mined areas.
  • the results of ExS were validated based on conventional field mines detection provided by Norway group in Angola and Israeli government for Golan Heights regions.
  • the total accuracy of presented method is very high (up to 90%) that provides a robust, generic and fast tool for possible minefields detection using remote sensing technique.
  • FIG. 18 schematically depicting the remote sensing of a region of interest (ROI).
  • the ROI 300 is suspected of being a minefield 330 in which a relatively large number of mines is hidden.
  • the mines are not generally visible or detectable by remote sensing per se.
  • the ROI is inhabited by biological organism, both inside and outside of the minefield.
  • the organisms are marked in this figure by small circles, one of which is marked by 200. They may be plants, bacteria, herbivores feeding of plants, or parasites of plants, or any other life-form, and they are detectable by remote sensing.
  • Radiation 410 emerges from the ROI and reaches a sensing unit 400.
  • the radiation is typically electromagnetic, and it is characterized by a range of wavelength to which the atmosphere is transparent, for example visible light and infra-red.
  • the radiation is typically sun-light reflected from the soil and from the organisms living on it.
  • the remote sensor typically comprises an airborne camera, typically mounted on an airplane, glider or balloon. Alternatively it may be mounted on any vehicle traveling on land.
  • the sensing unit produces an image 420 of the ROI.
  • the image comprises an array of image elements, each element detecting radiation emerging from a specific sub-region of the ROI.
  • Element 430 detects radiation from region 340, which in this figure is shown to be inside minefield 330, but may as well be located elsewhere in the ROI.
  • Element 430 comprises a description of detected intensity versus wavelength, which is described in reference to figure 19.
  • the intensity versus wavelength is a schematic representation of the MSSF from which the MSSD is calculated for the classification procedure. ;
  • FIG 19 schematically depicting a description of detected intensity versus wavelength. It is shown in this figure as a graph describing intensity T as a function of wavelength ' ⁇ '.
  • the bold line relates to a sub-region free of mines (control group), for example, and the dotted line relates to a sub-region within a minefield.
  • the function which is described by the dotted line refers to the CGST function, and the bold line relates to the MSSD function.
  • the two function are substantially indistinguishable at one wavelength ' ⁇ ', in this example, and are distinguishable (i.e. their difference is larger then the expected noise, not shown) at another wavelength ' ⁇ .
  • the intensity at ⁇ 1 may serve as a property of the description of intensity vs. wavelength, which is useful for minefield detection.
  • Sensing unit 400 of radiation 410 has been discussed in reference to figure 18.
  • analyzing unit 120 which receives information from sensing unit 400 in order to produce a form of identification 190 of minefields.
  • This identification may comprise a list of geographical coordinates, or it may comprise a map as depicted. The map distinguished between parts of the ROI that are relatively free of mines and parts that are considered to be minefields.
  • Map 190 is produced also by analyzing unit 120 by analyzing the image produced by sensing unit 400. Analyzing unit 110 receives the descriptions of intensity vs.
  • Analyzing unit 120 may be implemented, for example, using a general purpose suitably programmed computer, as will be described in further detail herein below.
  • analyzing unit 120 further receives a pre-existing representation 350 of the ROI. This may comprise a map, a photograph, a drawing or a picture of the ROI, or it may comprise a list of coordinates of objects of interest within the ROI. Analyzing unit 120 superimposes its indication of minefields on top of the provided pre-existing representation to produce a merged output, as schematically depicted in this figure.
  • analyzing unit 120 further receives a pre-existing representation 390 of the ROI.
  • This comprises auxiliary information concerning the location of minefields in the ROI.
  • the auxiliary information may be provided either by any of the mine-detection methods known in the art, or by a previous application of the present invention.
  • the purpose of this arrangement is for the present invention to augment a preexisting partial indication of the location of minefields, and for the present invention to perform a process of learning.
  • FIG 23 schematically depicting means for identifying a property of a description of intensity versus wavelength - analyzing unit 120 described in reference to figure 20.
  • the network may be constructed of analog means such as analog electrical circuits, and it may be simulated on a general purpose digital computer.
  • Figure 23 schematically shows one possible neural network design, of many that are well known in the art. This design employs three layers on (simulated) neurons. Layer 710 comprises a relatively large number of neurons, each receiving a signal proportional to the average intensity around a specific wavelength'.
  • the first neuron receives a signal proportional to the intensity of detected radiation around wavelength ' ⁇ ', the next - ' ⁇ , etc. Together, the neurons essential span the detected range of wavelengths. Each neuron emits a signal received by many if not all neurons in the next layer.
  • One middle layer 720 is depicted in figure 23, while several layer may be employed in cascade. Neurons in a middle layer essential behave like neurons in the first layer.
  • the last layer 730 comprises only one neuron, which emits one binary signal: minefield detected or not. As known in the art, each layer weights and then sums its inputs, employs a non-linear function such as a sigmoid, and emits on output.
  • Analyzing unit 110 comprises, according to this implementation of the present invention, at least one such network, which may be employed for the entire image elements, and alternatively, processing time may be accelerated by employing several identical networks.
  • the training of the neural network may take advantage of auxiliary information as described in reference to figure 22.
  • the training of the identification analyzing unit 120 is executed in two steps.
  • the first step ignores any information concerning minefields, and trains the identification block to distinguish between various types of soil and types of vegetation found in the ROI.
  • the second step takes any given information concerning minefields into account, and train networks to recognize them, networks for each type of soil or vegetation. Therefore, many differently trained networks are employed for a given ROI to identify minefields therein.
  • FIG 24 schematically depicting means for analyzing an image - analyzing unit 120 described in reference to figure 24. Since a property identifying a minefield image element has already been found analyzing unit 120, this block need only perform a segmentation of an image into two categories.
  • a simple implementation uses the output of a suitably trained neural network as described in reference to figure 23, essentially without modification.
  • a preferred implementation is depicted in figure 24 that improves the accuracy of minefield detection using the assumption that minefield tend to be contiguous geographic areas.
  • Layer 730 comprises re-enforcing interconnection among neighboring elements, so that each element depends in its neighbors.
  • the axes X and Y represent direction such as east-west and north-south.
  • a neuron is preferably required per each element in the image of the ROI.
  • the central neuron after weighting and summing its input from the previous layer 720, is uncertain of its decision, but input from most of its neighbors indicate that they would classify their sub-regions as minefield, in this case the central neuron can add this information to its sums, and join its neighbors in their decision.

Landscapes

  • Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

L'invention concerne un système d'identification et de localisation d'un champ de mines dans une région géographique présentant un intérêt (ROI), ledit système comprenant : c. une unité de détection pour recevoir le rayonnement électromagnétique de ladite ROI et générer au moins une image dudit rayonnement électromagnétique de ladite ROI, ladite ROI étant caractérisée par une localisation géographique spécifique ; et d. une unité d'analyse conçue pour classer la probabilité que ladite ou lesdites images soient un champ de mines à l'aide d'un ou de plusieurs paramètres spectraux associés à au moins une chimie exposée de ladite ROI ; lesdits paramètres spectraux étant sélectionnés dans le groupe consistant en SR, NDVI, PVI, SAVI, NLI, RDVI, MSR, WDVI, MNLI, NDVI*SR, SAVI*SR, TSAVI, rapport de vert, paramètres de rouge et de proche infrarouge, pente NIR-SWIR1, algorithme SAM en SWIR2, PPI, indice spectral d'azote, et n'importe quelle combinaison de ceux-ci.
PCT/IL2011/000872 2010-11-11 2011-11-10 Système et procédé de détection de champs de mines WO2012063241A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
ZA2013/03377A ZA201303377B (en) 2010-11-11 2013-05-09 System and method for detection of minefileds

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US41241610P 2010-11-11 2010-11-11
US61/412,416 2010-11-11

Publications (1)

Publication Number Publication Date
WO2012063241A1 true WO2012063241A1 (fr) 2012-05-18

Family

ID=46050456

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IL2011/000872 WO2012063241A1 (fr) 2010-11-11 2011-11-10 Système et procédé de détection de champs de mines

Country Status (2)

Country Link
WO (1) WO2012063241A1 (fr)
ZA (1) ZA201303377B (fr)

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103196838A (zh) * 2013-03-15 2013-07-10 辽宁师范大学 一种海岸河口富营养化高光谱遥感监测方法
CN104463836A (zh) * 2014-09-03 2015-03-25 中国科学院遥感与数字地球研究所 一种基于移动窗口的城市绿色空间遥感度量方法
CN105445229A (zh) * 2015-12-15 2016-03-30 河南农业大学 基于氮平衡光谱法的小麦春季氮肥施用方法及其追氮量模型的构建方法
CN105651679A (zh) * 2016-02-04 2016-06-08 华中农业大学 一种基于高光谱成像技术对培养基上的细菌菌落进行快速分类的方法
WO2016203689A1 (fr) * 2015-06-19 2016-12-22 パナソニックＩｐマネジメント株式会社 Dispositif de capture d'images
CN106682624A (zh) * 2016-12-29 2017-05-17 中国科学院深圳先进技术研究院 基于时间序列遥感信息的建成区提取方法及装置
CN107314990A (zh) * 2017-08-28 2017-11-03 北京师范大学 一种春玉米遥感识别方法
CN109638871A (zh) * 2019-01-28 2019-04-16 东北电力大学 考虑风电接入的大规模交直流混联系统主网划分方法
CN109799193A (zh) * 2019-02-19 2019-05-24 北京英视睿达科技有限公司 污染分布立体监测方法及系统
CN110705655A (zh) * 2019-11-05 2020-01-17 云南省烟草农业科学研究院 一种基于光谱和机器视觉耦合的烟叶分类方法
WO2020068694A1 (fr) 2018-09-25 2020-04-02 Urugus S.A. Agents transducteurs et dispositifs de détection à distance
CN111292325A (zh) * 2020-03-31 2020-06-16 中国地质调查局西安地质调查中心(西北地质科技创新中心) 一种基于遥感技术的萤石矿识别方法及系统
CN111325126A (zh) * 2020-02-12 2020-06-23 二十一世纪空间技术应用股份有限公司 遥感图像的解译方法及装置
CN111678871A (zh) * 2020-06-16 2020-09-18 中国气象科学研究院 一种非生长季植被覆盖度遥感估算方法
WO2020250214A1 (fr) * 2019-06-12 2020-12-17 4M Analytics Ltd Système et procédé de génération d'une carte de champ d'engins explosifs
CN112683822A (zh) * 2020-12-17 2021-04-20 淮阴师范学院 基于可见光波段的植被与非植被识别方法
US20210383546A1 (en) * 2018-10-04 2021-12-09 Nec Corporation Learning device, image processing device, learning method, image processing method, learning program, and image processing program
US11519781B2 (en) * 2016-07-22 2022-12-06 Nec Corporation Image processing device, image processing method, and recording medium
CN117035259A (zh) * 2023-06-21 2023-11-10 北京市遥感信息研究所 基于多层神经网络的任务与卫星平台关联映射方法和系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5942899A (en) * 1997-07-28 1999-08-24 Aerojet-General Corporation Hyperspectral radiometric mine detector based upon sensing microwave brightness temperature and interference contrast signatures
US6473025B2 (en) * 2001-03-28 2002-10-29 Stolar Horizon Landmine locating system
US20060113998A1 (en) * 2002-11-22 2006-06-01 National Institute For Materials Science Mine detector with nqr-squid
US20080002185A1 (en) * 2006-06-08 2008-01-03 Anatoly Gitelson System and methods for non-destructive analysis
US20090058710A1 (en) * 2006-05-09 2009-03-05 Levitan Arthur C Methods and apparatus for detecting threats using radar
US20100104191A1 (en) * 2007-03-26 2010-04-29 Mcgwire Kenneth C Data analysis process
US20100223276A1 (en) * 2007-03-27 2010-09-02 Faleh Jassem Al-Shameri Automated Generation of Metadata for Mining Image and Text Data
US20110242328A1 (en) * 2010-04-05 2011-10-06 Lockheed Martin Corporation Ir-based multispectral disturbed ground detection

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5942899A (en) * 1997-07-28 1999-08-24 Aerojet-General Corporation Hyperspectral radiometric mine detector based upon sensing microwave brightness temperature and interference contrast signatures
US6473025B2 (en) * 2001-03-28 2002-10-29 Stolar Horizon Landmine locating system
US20060113998A1 (en) * 2002-11-22 2006-06-01 National Institute For Materials Science Mine detector with nqr-squid
US20090058710A1 (en) * 2006-05-09 2009-03-05 Levitan Arthur C Methods and apparatus for detecting threats using radar
US20080002185A1 (en) * 2006-06-08 2008-01-03 Anatoly Gitelson System and methods for non-destructive analysis
US20100104191A1 (en) * 2007-03-26 2010-04-29 Mcgwire Kenneth C Data analysis process
US20100223276A1 (en) * 2007-03-27 2010-09-02 Faleh Jassem Al-Shameri Automated Generation of Metadata for Mining Image and Text Data
US20110242328A1 (en) * 2010-04-05 2011-10-06 Lockheed Martin Corporation Ir-based multispectral disturbed ground detection

Cited By (28)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103196838A (zh) * 2013-03-15 2013-07-10 辽宁师范大学 一种海岸河口富营养化高光谱遥感监测方法
CN104463836A (zh) * 2014-09-03 2015-03-25 中国科学院遥感与数字地球研究所 一种基于移动窗口的城市绿色空间遥感度量方法
WO2016203689A1 (fr) * 2015-06-19 2016-12-22 パナソニックＩｐマネジメント株式会社 Dispositif de capture d'images
CN105445229A (zh) * 2015-12-15 2016-03-30 河南农业大学 基于氮平衡光谱法的小麦春季氮肥施用方法及其追氮量模型的构建方法
CN105445229B (zh) * 2015-12-15 2018-06-05 河南农业大学 基于氮平衡光谱法的小麦春季氮肥施用方法及其追氮量模型的构建方法
CN105651679A (zh) * 2016-02-04 2016-06-08 华中农业大学 一种基于高光谱成像技术对培养基上的细菌菌落进行快速分类的方法
US11519781B2 (en) * 2016-07-22 2022-12-06 Nec Corporation Image processing device, image processing method, and recording medium
CN106682624B (zh) * 2016-12-29 2019-08-02 中国科学院深圳先进技术研究院 基于时间序列遥感信息的建成区提取方法及装置
CN106682624A (zh) * 2016-12-29 2017-05-17 中国科学院深圳先进技术研究院 基于时间序列遥感信息的建成区提取方法及装置
CN107314990A (zh) * 2017-08-28 2017-11-03 北京师范大学 一种春玉米遥感识别方法
CN107314990B (zh) * 2017-08-28 2020-01-17 北京师范大学 一种春玉米遥感识别方法
EP3844483A4 (fr) * 2018-09-25 2022-05-18 Urugus S.A. Agents transducteurs et dispositifs de détection à distance
WO2020068694A1 (fr) 2018-09-25 2020-04-02 Urugus S.A. Agents transducteurs et dispositifs de détection à distance
US20210383546A1 (en) * 2018-10-04 2021-12-09 Nec Corporation Learning device, image processing device, learning method, image processing method, learning program, and image processing program
CN109638871B (zh) * 2019-01-28 2022-04-15 东北电力大学 考虑风电接入的大规模交直流混联系统主网划分方法
CN109638871A (zh) * 2019-01-28 2019-04-16 东北电力大学 考虑风电接入的大规模交直流混联系统主网划分方法
CN109799193A (zh) * 2019-02-19 2019-05-24 北京英视睿达科技有限公司 污染分布立体监测方法及系统
US11885594B2 (en) 2019-06-12 2024-01-30 4M Analytics Ltd. System and method of generating explosive devices field map
WO2020250214A1 (fr) * 2019-06-12 2020-12-17 4M Analytics Ltd Système et procédé de génération d'une carte de champ d'engins explosifs
CN110705655A (zh) * 2019-11-05 2020-01-17 云南省烟草农业科学研究院 一种基于光谱和机器视觉耦合的烟叶分类方法
CN111325126B (zh) * 2020-02-12 2023-11-03 二十一世纪空间技术应用股份有限公司 遥感图像的解译方法及装置
CN111325126A (zh) * 2020-02-12 2020-06-23 二十一世纪空间技术应用股份有限公司 遥感图像的解译方法及装置
CN111292325A (zh) * 2020-03-31 2020-06-16 中国地质调查局西安地质调查中心(西北地质科技创新中心) 一种基于遥感技术的萤石矿识别方法及系统
CN111292325B (zh) * 2020-03-31 2021-01-12 中国地质调查局西安地质调查中心(西北地质科技创新中心) 一种基于遥感技术的萤石矿识别方法
CN111678871A (zh) * 2020-06-16 2020-09-18 中国气象科学研究院 一种非生长季植被覆盖度遥感估算方法
CN112683822A (zh) * 2020-12-17 2021-04-20 淮阴师范学院 基于可见光波段的植被与非植被识别方法
CN112683822B (zh) * 2020-12-17 2024-04-05 淮阴师范学院 基于可见光波段的植被与非植被识别方法
CN117035259A (zh) * 2023-06-21 2023-11-10 北京市遥感信息研究所 基于多层神经网络的任务与卫星平台关联映射方法和系统

Also Published As

Publication number Publication date
ZA201303377B (en) 2014-07-30

Similar Documents

Publication Publication Date Title
WO2012063241A1 (fr) Système et procédé de détection de champs de mines
Jia et al. Status and application of advanced airborne hyperspectral imaging technology: A review
Adiri et al. Recent advances in the use of public domain satellite imagery for mineral exploration: A review of Landsat-8 and Sentinel-2 applications
Pu Hyperspectral remote sensing: fundamentals and practices
Thenkabail et al. Hyperspectral remote sensing of vegetation and agricultural crops
Clark et al. Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales
Han et al. Mapping hydrothermally altered rocks with Landsat 8 imagery: A case study in the KSM and Snow field zones, northwestern British Columbia
Ashraf et al. Introduction to remote sensing of biomass
Nyamekye et al. Mapping changes in artisanal and small-scale mining (ASM) landscape using machine and deep learning algorithms.-a proxy evaluation of the 2017 ban on ASM in Ghana
Khaleghi et al. Spectral angle mapping, spectral information divergence, and principal component analysis of the ASTER SWIR data for exploration of porphyry copper mineralization in the Sarduiyeh area, Kerman province, Iran
Gürsoy et al. Detecting of lithological units by using terrestrial spectral data and remote sensing image
Lammoglia et al. Unraveling hydrocarbon microseepages in onshore basins using spectral–spatial processing of advanced spaceborne thermal emission and reflection radiometer (ASTER) data
Thenkabail et al. Fundamentals, sensor systems, spectral libraries, and data mining for vegetation
Weingarten et al. Early detection of a tree pathogen using airborne remote sensing
Pandey et al. The Use of Hyperspectral Earth Observation Data for Land Use/Cover ClassificationPresent Status, Challenges, and Future Outlook
Milewski et al. Analyses of the impact of soil conditions and soil degradation on vegetation vitality and crop productivity based on airborne hyperspectral VNIR–SWIR–TIR data in a Semi-Arid Rainfed Agricultural Area (Camarena, Central Spain)
Varshney et al. Hyperspectral sensors and applications
Oliveira et al. Advances in hyperspectral sensing in agriculture: a review
Tuominen et al. Detection of environmental change using hyperspectral remote sensing at Olkiluoto repository site
Bajić et al. Developing a hyperspectral non-technical survey for minefields via UAV and helicopter
Patterson et al. Integrated sensor systems for UAS
Jadhav et al. Hyperspectral remote sensing for agricultural management: a survey
Xing et al. Development and application of hyperspectral remote sensing
Chen et al. Multimodal Remote sensing science and technology
Mishra et al. Spectral indices across remote sensing platforms and sensors relating to the three poles: An overview of applications, challenges, and future prospects

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11839472

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11839472

Country of ref document: EP

Kind code of ref document: A1