CN105426881A - Mountain background thermal field model constrained underground heat source daytime remote sensing detection locating method - Google Patents

Abstract

The invention discloses a mountain background thermal field model constrained underground heat source daytime remote sensing detection locating method, belongs to the field of intersection of remote sensing technology, natural geography and mode identification, and aims at performing thermal field simulation of a mountain background to obtain a mountain background thermal field model and performing background filtering on an original remote sensing infrared image with the thermal field model acting as the constraint so as to clearly reveal the location of an underground target. The mountain background thermal field model constrained underground heat source daytime remote sensing detection locating method includes a mountain background thermal field model establishment step, a mountain background thermal field 8-bit-to-16-bit mapping step, a background filtering step by utilizing the mountain background thermal field model after mapping, and an underground target space constraint mean clustering detection locating step. The model is constructed by utilizing the mountain background thermal field, and gray scale mapping is performed on the model by utilizing the real mountain background so that the established thermal field model is ensured to be close to the real mountain background thermal field, and finally background filtering processing is performed on the real remote sensing infrared image and the location of the underground target is determined.

Description

Remote sensing localization method between underground thermal source daytime of massif background heat field model constraint

Technical field

The invention belongs to the crossing domain of remote sensing technology, physical geography and pattern-recognition, remote sensing localization method between the underground thermal source daytime being specifically related to the constraint of a kind of massif background heat field model.

Background technology

The development of the mankind be unable to do without various natural resources, in order to obtain abundant subterranean resource, mineral resources and groundwater resource etc., needs to build a large amount of underground installations, and thus the Detection Techniques of the buried target such as hypogee, underground installation seem and become more and more important.All temperature can produce heat radiation higher than the object of absolute zero, buried target and peripheral region have different thermodynamic behaviours, under the effect of external environment, the existence of buried target can affect the internal heat conductive process of peripheral region, causes the place that there is buried target and ambient background to occur time dependent temperature contrast.From statistical significance, namely the temperature field of buried target is higher or lower than the temperature field of massif background.Therefore, these buried targets can be regarded as the underground thermal source being different from background thermal source.Compared to other detection means, the detection of infrared technique means has certain advantage.Non-remote sensing means cannot realize large-area simultaneous observation, are difficult to use under rugged environment, and the speed of obtaining information is slow, and time and effort consuming.Existing conventional remote sensing means are mainly for the conditional object on earth's surface or the water surface, and obtaining information is subject to the restriction of physical environment, background environment, cannot detect deeper subsurface thermal source target.At present, electromagnetically induced technology, only to detect the metal target under shallow layer surface, is easily subject to underground simultaneously and is scattered the impact of metal fragment.Therefore, the detection of infrared technique means becomes a kind of effective underground objects detection means.

At present, certain research is also had to the detection of buried target both at home and abroad.The domestic detection mainly concentrating on shallow-layer target, and for the detection of target under multi-temporal image, and these buried targets are the underground thermal source of large scale mostly.Also the relevant report of deep layer (being greater than 10m apart from surface distance) underground thermal source target detection is had no, the especially underground thermal source target of small scale in multinational.There is the research utilizing airborne medium wave and long-wave infrared to sweep sensor Underground target abroad, but have no the relevant report utilizing remote sensing images to carry out underground objects detection.For the detection of the Band object of plane, existing mode identification method does not adopt the method for background filtering.And it detects is only suspected target district, not accurate buried target position, location.Moreover, the false dismissed rate detecting the suspected target district obtained is undesirable, and false alarm rate is high, and positional accuracy is not high yet.

Summary of the invention

Under the present invention proposes the constraint of a kind of massif background heat field model, to the method for Detection location between the distributed thermal source in underground daytime of deep layer (being greater than 10m apart from surface distance), solve existing the problem for shallow underground thermal source Detection location, simulation softwares is utilized to carry out the analog simulation of thermal field to massif, analyze the thermal field model obtaining massif body background, true infrared figure is utilized to carry out mapping to massif thermal field model unified, and utilize the model after mapping to carry out background filtering process to true infrared image, reduce massif body background thermal field to the impact of detection, the position at buried target place, last accurately location.

The invention provides detecting and positioning method between a kind of distributed thermal source in underground daytime based on the constraint of massif background heat field model.When setting up massif background heat field model, do suitable simplification according to actual massif thermal field, the model be simplified, concrete steps are as follows:

(1) foundation of massif background heat field model, comprises following sub-step:

(1.1) establishment step of mountain model

(1.1.1) obtained the 3-dimensional digital elevation model of massif by remote sensing mapping, obtain the altitude data information of true massif.

(1.1.2) structure of ANSYS geometry mountain model is formed with point, line, surface and body, and point (coordinate) is the basis building geometric model.So according to above-mentioned altitude information, the establishment of geometry massif generates closed curve by key point, and closed curve generates plane, then surrounds geometry massif by the curved surface closed.And geometry massif just constitutes whole mountain model.

(1.2) the FEM meshing step of mountain model

Anasy FEM meshing carries out the vital step of numerical simulation analysis, model due to massif is not rule, therefore free mesh is carried out, be free to automatically generate triangle or tetrahedral grid on the whole, body generates tetrahedral grid automatically, manually carries out the control of smart dimensions simultaneously.

(1.3) mountain model boundary condition is arranged and solution procedure

Massif after above-mentioned stress and strain model is carried out to the setting of load boundary condition, the elementary heat of massif heat transfer and massif-air thermal convection is utilized to transmit physical basis, the parameter of heat trnasfer is set according to the pyroconductivity K of massif and the thermal convection rate Φ of massif-air, the thermo parameters method calculating massif is solved through Ansys, finally the mapping of gray-scale map and the adjustment of temperature resolution are carried out to the thermo parameters method of massif, finally obtain massif thermal field model.

(2) massif background thermal field 8 is to 16 replacement response, comprises following sub-step:

Because the massif background infra-red heat field model image set up is 8, and real massif background heat field picture is 16, so need mapping process massif background heat field model obtained above and true massif background thermal field being carried out to 8 to 16.When the thermal field of model and real heterogeneity phantom change roughly the same, ensure that model thermal field is more close to the thermal field of true massif.

(2.1) determining step of true massif background gray levels scope

The thermal field of true massif background is subject to the impact of many extraneous factors, such as house, road etc.The ratio shared in whole background due to these extraneous factors is very little, therefore can be used as interference.So need to carry out range constraint process, by statistics with histogram, carry out threshold process, reject the interference of extraneous factor, detailed process is as follows:

(2.1.1) establish pixel grayscale in variable r representative image, under discrete situation, use r _krepresent discrete gray levels, with P (r _k) represent probability density function, have following formula to set up:

$\begin{array}{cc}{P}_r\left(r_k\right)=\frac{n_k}{n}& (0\≤r_k\≤l)\end{array}$

k＝0,1,2...l-1

N in formula _kfor there is r in image _kthe pixel count of gray scale, n is pixel count sum in image, be exactly the frequency in theory of probability, l is the total number of gray level.

Known extraneous factor impact accounts for area occupied in entire image than being P%, then has following formula:

P _r(r _k)≥P

Accumulative grey level histogram successively, if aggregate-value is more than or equal to object proportion, stops cumulative, record r _kvalue, standard as a setting.

(2.2) the map correction step of massif thermal field model

The gray-scale value of massif thermal field Model Background carries out linear mapping correction process according to the minimum gray value of value obtained above and image.Concrete formula is as follows:

$O=\frac{I-I_l}{I_h-I_l}(O_h-O_l)+O_l$

Wherein, I is the gray-scale value of thermal field model, I _lfor the minimum brightness gray-scale value of thermal field model, I _hfor the maximum brightness gray-scale value of thermal field model, O _lfor the minimum brightness gray-scale value of true massif infrared image, O _hfor the above-mentioned r tried to achieve _k, O is the massif background model after map correction.

(3) the massif background heat field model after mapping is utilized to carry out background filter step

Due to the angle of remote sensor observation and the impact of solar illumination angle, massif can be divided into sunny side and the back.The massif part of sun direct irradiation is called sunny side, and the sun part of direct irradiation cannot be called the back.Massif background and surrounding environment have heat radiation, and heat radiation when direct irradiation and non-immediate are irradiated there are differences.Moreover, between daytime massif background thermal field and night massif background thermal field also different, the present invention is directed Detection location between underground thermal source daytime of massif background heat field model constraint.First find the pixel of the boundary of the back and sunny side, utilize the method matching back of least square fitting and the cut-off rule of sunny side, then grey level compensation operation is carried out to the back, finally carry out background filtering process.

(3.1) original infrared image back sunny side separatrix extraction step

(3.1.1) difference of the back and sunny side gray scale difference value is utilized to determine the pixel of boundary, it is east-west for can defining boundaries according to the information of view data, so only need more upper and lower pixel gray-scale value, if pixel grey scale difference contiguous is up and down greater than K, formula is as follows:

G(x,y)>G(x,y-1)+K

G(x,y)>G(x,y-2)+K

G(x,y)>G(x,y-3)+K

G(x,y)>G(x,y-1)+K

If above-mentioned four inequality are set up, just (x, y) is considered as the pixel near separatrix, traversal full figure obtains all pixels near separatrix.

(3.1.2) next carry out cubic polynomial least square fitting linear process to the pixel near all separatrix obtained above, detailed process is as follows:

Wherein for the cubic polynomial of least square fitting, err is error objective function, and by making, err is minimum reaches optimum cubic polynomial matching, obtains final separatrix.

(3.2) step of sunny side sun direct irradiation energy is eliminated

After obtaining infrared image back region and sunny side region according to separatrix, the gray-scale value of following mapping policy to infrared image infrared image sunny side region is adopted to eliminate:

${F^{\′}}_{noshadow}(i,j)=A\[m_{shadow}+\frac{F_{noshadow}(i,j)-m_{noshadow}}{{\mathrm{\σ}}_{noshadow}}*{\mathrm{\σ}}_{shadow}\]$

D＝F _n' _oshadow(i,j)-F _noshadow(i,j)

F' in formula _noshadow(i, j) is the sunny side area grayscale value of the non-direct irradiation of the infrared image sun, F _noshadow(i, j) is infrared image sun direct irradiation sunny side area grayscale value, m _shadowand σ _shadowaverage and the variance of infrared image sunny side area grayscale value, m _noshadowand σ _noshadowbe average and the variance of contiguous non-infrared image infrared image back area grayscale value, A is compensation intensity coefficient, and D is the energy gray-scale value of sun direct irradiation sunny side.

Obtain the gray-scale value of sun direct irradiation impact from above-mentioned after, travel through all the moon detected

Region, face, gray-scale value infrared image sunny side region being deducted each point of infrared figure compensates D, and be eliminated the gray-scale map after shining upon impact.Do not consider the impact shined upon before, filtered by the component shined upon, filter sun sunny side by the energy shined upon, shine upon the thermal caused, absorbed by sunny side, under thermal equilibrium condition, filtering is by the component of sun direct irradiation.

(3.3) the true infrared figure background filtering treatment step of massif

Utilize the massif background heat field model emulating and obtain as the background of the infrared figure of true massif, because the target heat radiation in massif meets heat conducting mathematical model, suppose at certain a moment t ₀target location (x ₀, y ₀, z ₀) heat radiation curved surface be BT (x, y, z, t)

BT(x,y,z,t)

＝B(x,y,z,t)+T(x,y,z,t)*R _b(x,y,z,t)

+A(x,y,z,t)+δ(x,y,z,t)

This moment target background body radiation field BT (x, y, z, t), by multi-dielectric body radiation field B (x, y, z, t), by objective body radiation field T (x, y, z, t) the * R that multi-dielectric body has distorted _b(x, y, z, t), the radiant quantity δ (x, y, z, t) that water body/terrain and air contact surfaces scatter and disappear and shine upon and affect A (x, y, z, t) and jointly produce.

The impact of target background body radiation field produces primarily of B (x, y, z, t) and A (x, y, z, t), therefore has following formula:

T(x,y,z,t)＝BT(x,y,z,t)-k*B(x,y,z,t)-A(x,y,z,t)

Wherein T (x, y, z, t) be the approximate thermal radiation field of target, BT (x, y, z, t) be target background body radiation field, B (x, y, z, t) be multi-dielectric body background radiation field, A (x, y, z, t) for shining upon the energy field of impact, k is the adjustability coefficients of background radiation field.Finally obtain the filtered image of true massif infrared figure background.

(4) buried target space constraint mean cluster Detection location step

From the filtered image of background, choosing image block to be identified has s, and template size is 3*3, is b respectively ₁, b ₂, b ₃..., b _s, have object region and lower driftlessness image-region under comprising, when selection, avoid the impact of house and road.

Utilize space constraint means clustering algorithm road segment segment b ₁, b ₂, b ₃..., b _sobject region and lower driftlessness image-region two class is had under being divided into.The specific implementation process of space constraint means clustering algorithm is as follows:

Step1: for all sample point b _i, calculate distance than

$V_i=\underset{j=1}{\overset{s}{\Σ}}\frac{d(b_i,b_j)}{\underset{l=1}{\overset{s}{\Σ}}d(b_l,b_j)}$

Select V _iminimum some b _ias first class heart, juxtaposition q=1;

Step2: to p=1,2, by b _i, i=1,2 ..., s is assigned to from its nearest class, and upgrades the class heart i=1,2, N _iit is the sample number of the i-th class;

Step3: put q=q+1, if q>2, algorithm stops;

Step4: select the best initial center point of next class for making minimum some b _i, proceed to Step2.

Obtain the result after cluster by above formula, large that class of gray-scale value is as a class of doubtful buried target, and the little class of gray-scale value is as a class of non-doubtful buried target.The position of buried target is obtained finally by space constraint clustering algorithm.

Accompanying drawing explanation

Fig. 1 is schematic flow sheet of the present invention;

Fig. 2 (a) is massif altitude information schematic diagram;

Fig. 2 (b) is massif level line schematic diagram;

Fig. 2 (c) is schematic diagram before massif stress and strain model;

Fig. 2 (d) is schematic diagram after massif stress and strain model;

Fig. 3 is the radiation model schematic diagram of massif;

Fig. 4 is the true infrared image of massif;

Fig. 5 is the background heat field model image after mapping;

Fig. 6 is pixel schematic diagram near massif male and female face separatrix;

Fig. 7 is massif male and female face separatrix schematic diagram;

Fig. 8 is the infrared image after massif eliminates the impact of sun direct irradiation;

Fig. 9 is based on the filtered infrared image of thermal field Model Background;

Figure 10 is underground objects detection result schematic diagram.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.In addition, if below in described each embodiment of the present invention involved technical characteristic do not form conflict each other and just can mutually combine.

As shown in Figure 1, wherein concrete implementation method comprises the following steps flow process of the present invention.The massif background heat field model that the present invention includes after massif background heat field model establishment step, massif background thermal field 8 to 16 replacement response steps, utilization mapping carries out background filter step, buried target space constraint mean cluster Detection location step:

(1) foundation of massif background heat field model, comprises following sub-step:

(1.1) establishment step of mountain model

(1.1.1) obtained the 3-dimensional digital elevation model of massif by remote sensing mapping, obtain the altitude data information of true massif, as shown in Fig. 2 (a).

(1.1.2) structure of ANSYS geometry mountain model is formed with point, line, surface and body, and point (coordinate) is the basis building geometric model.So according to above-mentioned altitude information, the establishment of geometry massif generates closed curve by key point, and closed curve generates plane, then surrounds geometry massif by the curved surface closed.And geometry massif just constitutes whole mountain model, as shown in Fig. 2 (b) He Fig. 2 (c).

(1.2) the FEM meshing step of mountain model

Anasy FEM meshing carries out the vital step of numerical simulation analysis, model due to massif is not rule, therefore free mesh is carried out, be free to automatically generate triangle or tetrahedral grid on the whole, body generates tetrahedral grid automatically, manually carries out the control of smart dimensions simultaneously.This example adopts secondary tetrahedron element (No. 92 unit), ensures to calculate computational accuracy, as shown in Fig. 2 (d).

(1.3) mountain model boundary condition is arranged and solution procedure

Massif after above-mentioned stress and strain model is carried out to the setting of load boundary condition, the elementary heat of massif heat transfer and massif-air thermal convection is utilized to transmit physical basis, the parameter of heat trnasfer is set according to the pyroconductivity K of massif and the thermal convection rate Φ of massif-air, the thermo parameters method calculating massif is solved through Ansys, finally the mapping of gray-scale map and the adjustment of temperature resolution are carried out to the thermo parameters method of massif, finally obtain massif radiation model.In this example, K=3.49Kg/m3, Φ=3W/ (m^2.C), solving result as shown in Figure 3.

(2) massif background thermal field 8 is to 16 replacement response processes, comprises following sub-step:

Because the massif background infra-red heat field model image set up is 8, and real massif background heat field picture is 16, so need mapping process massif background heat field model obtained above and true massif background thermal field being carried out to 8 to 16.When the thermal field of model and real heterogeneity phantom change roughly the same, ensure that model thermal field is more close to the thermal field of true massif.

(2.1) determining step of the scope of true massif background gray levels

The thermal field of true massif background is subject to the impact of many extraneous factors, such as house, road etc.The ratio shared in whole background due to these extraneous factors is very little, therefore can be used as interference.So need to carry out range constraint process, by statistics with histogram, carry out threshold process, reject the interference of extraneous factor, as shown in Figure 4, detailed process is as follows for true massif infrared image:

(2.1.1) establish pixel grayscale in variable r representative image, under discrete situation, use r _krepresent discrete gray levels, with P (r _k) represent probability density function, have following formula to set up:

$\begin{array}{cc}{P}_r\left(r_k\right)=\frac{n_k}{n}& (0\≤r_k\≤l)\end{array}$

k＝0,1,2...l-1

N in formula _kfor there is r in image _kthe pixel count of gray scale, n is pixel count sum in image, be exactly the frequency in theory of probability, l is the total number of gray level.

Known extraneous factor impact accounts for area occupied in entire image than being P%, then has following formula:

P _r(r _k)≥P

Accumulative grey level histogram successively, if aggregate-value is more than or equal to object proportion, stops cumulative, record r _kvalue, standard as a setting.In this example, P=0.5, r _k=31000.

(2.2) the map correction step of massif thermal field model

The gray-scale value of massif thermal field background carries out linear mapping correction process according to the minimum gray value of value obtained above and image.Concrete formula is as follows:

$O=\frac{I-I_l}{I_h-I_l}(O_h-O_l)+O_l$

Wherein, I is the gray-scale value of thermal field model, I _lfor the minimum brightness gray-scale value of thermal field model, I _hfor the maximum brightness gray-scale value of thermal field model, O _lfor the minimum brightness gray-scale value of true massif infrared image, O _hfor the above-mentioned r tried to achieve _k, O is the massif background model after map correction.Massif background after correction as shown in Figure 5.In this example, O _l=29852.

(3) utilize the massif background heat field model after mapping to carry out background filter step, comprise following sub-step:

Due to the angle of remote sensor observation and the impact of solar illumination angle, massif can be divided into sunny side and the back.The massif part of sun direct irradiation is called sunny side, and the sun part of direct irradiation cannot be called the back.Massif background and surrounding environment have heat radiation, and heat radiation when direct irradiation and non-immediate are irradiated there are differences.Moreover, between daytime massif background thermal field and night massif background thermal field also different, the present invention is directed Detection location between underground thermal source daytime of massif background heat field model constraint.

First find the pixel of the boundary of the back and sunny side, utilize the method matching back of least square fitting and the cut-off rule of sunny side, then grey level compensation operation is carried out to the back, finally carry out background filtering process.

(3.1) original infrared image back sunny side separatrix extraction step

(3.1.1) difference of the back and sunny side gray scale difference value is utilized to determine the pixel of boundary, it is east-west for can defining boundaries according to the information of view data, so only need more upper and lower pixel gray-scale value, if pixel grey scale difference contiguous is up and down greater than K, formula is as follows:

G(x,y)>G(x,y-1)+K

G(x,y)>G(x,y-2)+K

G(x,y)>G(x,y-3)+K

G(x,y)>G(x,y-1)+K

If above-mentioned four inequality are set up, just (x, y) is considered as the pixel near separatrix, traversal full figure obtains all pixels near separatrix, as shown in Figure 6.In this example, K=40.

(3.1.2) next carry out cubic polynomial least square fitting linear process to the pixel near all separatrix obtained above, detailed process is as follows:

Wherein for the cubic polynomial of least square fitting, err is error objective function, and by making, err is minimum reaches optimum cubic polynomial matching, obtains final separatrix, as shown in Figure 7.In this example, a ₀=0, a ₁=0.001, a ₂=0.4635, a ₃=42.5124.

(3.2) step of sunny side sun direct irradiation energy is eliminated

After obtaining infrared image back region and sunny side region according to separatrix, the gray-scale value of following mapping policy to infrared image infrared image sunny side region is adopted to eliminate:

${F^{\′}}_{noshadow}(i,j)=A\[m_{shadow}+\frac{F_{noshadow}(i,j)-m_{noshadow}}{{\mathrm{\σ}}_{noshadow}}*{\mathrm{\σ}}_{shadow}\]$

D＝F _n' _oshadow(i,j)-F _noshadow(i,j)

F' in formula _noshadow(i, j) is the sunny side area grayscale value of the non-direct irradiation of the infrared image sun, F _noshadow(i, j) is infrared image sun direct irradiation sunny side area grayscale value, m _shadowand σ _shadowaverage and the variance of infrared image sunny side area grayscale value, m _noshadowand σ _noshadowbe average and the variance of contiguous non-infrared image infrared image back area grayscale value, A is compensation intensity coefficient, and D is the energy gray-scale value of sun direct irradiation sunny side.In this example, m _shadow=6478, m _noshadow=4478, σ _shadow=55.2752, σ _noshadow=61.9714, D=225, A=1.0.

Obtain the gray-scale value of sun direct irradiation generation from above-mentioned after, travel through all sun detected

Region, face, gray-scale value infrared image sunny side region being deducted each point of infrared figure compensates D, and be eliminated the gray-scale map after shining upon impact.Do not consider the impact shined upon before, the component shined upon is filtered, filter sun sunny side by the energy shined upon, shine upon the thermal caused, absorbed by sunny side, under thermal equilibrium condition, filtering is by the component of sun direct irradiation, and result as shown in Figure 8.

(3.3) the true infrared figure background filtering treatment step of massif

Utilize the massif background heat field model emulating and obtain as the background of the infrared figure of true massif, because the target heat radiation in massif meets heat conducting mathematical model, suppose at certain a moment t ₀target location (x ₀, y ₀, z ₀) heat radiation curved surface be BT (x, y, z, t)

BT(x,y,z,t)

＝B(x,y,z,t)+T(x,y,z,t)*R _b(x,y,z,t)

+A(x,y,z,t)+δ(x,y,z,t)

This moment target background body radiation field BT (x, y, z, t), by multi-dielectric body radiation field B (x, y, z, t), by objective body radiation field T (x, y, z, t) the * R that multi-dielectric body has distorted _b(x, y, z, t), the radiant quantity δ (x, y, z, t) that water body/terrain and air contact surfaces scatter and disappear and shine upon and affect A (x, y, z, t) and jointly produce.

The impact of target background body radiation field produces primarily of B (x, y, z, t) and A (x, y, z, t), therefore has following formula:

T(x,y,z,t)＝BT(x,y,z,t)-k*B(x,y,z,t)-A(x,y,z,t)

Wherein T (x, y, z, t) be the approximate thermal radiation field of target, BT (x, y, z, t) be target background body radiation field, B (x, y, z, t) be multi-dielectric body background radiation field, A (x, y, z, t) for shining upon the energy field of generation, k is the adjustability coefficients of background radiation field.Finally obtain the filtered image of true massif infrared figure background, as shown in Figure 9.In this example, k=0.8.

(4) buried target space constraint mean cluster Detection location step

From the filtered image of background, choosing image block to be identified has s, and template size is 3*3, is b respectively ₁, b ₂, b ₃... b _s, have object region and lower driftlessness image-region under comprising, when selection, avoid the impact of house and road.

Utilize space constraint means clustering algorithm road segment segment b ₁, b ₂, b ₃..., b _sobject region and lower driftlessness image-region two class is had under being divided into.The specific implementation process of space constraint means clustering algorithm is as follows:

Step1: for all sample point b _i, calculate distance than

$V_i=\underset{j=1}{\overset{s}{\Σ}}\frac{d(b_i,b_j)}{\underset{l=1}{\overset{s}{\Σ}}d(b_l,b_j)}$

Select V _iminimum some b _ias first class heart, juxtaposition q=1;

Step2: to p=1,2, by b _i, i=1,2 ..., s is assigned to from its nearest class, and upgrades the class heart i=1,2, N _iit is the sample number of the i-th class;

Step3: put q=q+1, if q>2, algorithm stops;

Step4: select the best initial center point of next class for making minimum some b _i, proceed to Step2.

In this example, m ₁=310, m ₂=400

Obtain the result after cluster by above formula, large that class of gray-scale value is as a class of doubtful buried target, and the little class of gray-scale value is as a class of non-doubtful buried target.The position of buried target is obtained, as shown in Figure 10 finally by space constraint clustering algorithm.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. a remote sensing localization method between underground thermal source daytime of massif background heat field model constraint, it is characterized in that, described method comprises the steps:

(1) foundation of massif background heat field model, comprises following sub-step:

(1.1) foundation of mountain model;

(1.2) FEM meshing of mountain model;

(1.3) mountain model boundary condition arranges and solves;

(2) massif background thermal field 8 is to 16 replacement response, comprises following sub-step:

(2.1) determination of true massif background gray levels scope;

(2.2) map correction of massif thermal field model;

(3) utilize the massif background heat field model after mapping to carry out background filtering, comprise following sub-step:

(3.1) original infrared image back sunny side separatrix is extracted;

(3.2) sunny side sun direct irradiation energy is eliminated;

(3.3) the true infrared figure background filtering process of massif;

(4) buried target space constraint mean cluster Detection location.

2. the method for claim 1, is characterized in that, described step (4) specifically comprises following sub-step:

From the filtered image of background, choosing image block to be identified has s, is b respectively ₁, b ₂, b ₃..., b _s, have object region and lower driftlessness image-region under comprising, utilize space constraint means clustering algorithm road segment segment b ₁, b ₂, b ₃..., b _shave object region and lower driftlessness image-region two class under being divided into, wherein the specific implementation process of space constraint means clustering algorithm is as follows:

Step1: for all sample point b _i, calculate distance than

$V_i=\underset{j=1}{\overset{s}{\Σ}}\frac{d(b_i,b_j)}{\underset{l=1}{\overset{s}{\Σ}}d(b_l,b_j)}$

Select V _iminimum some b _ias first class heart, juxtaposition q=1;

Step2: to p=1,2, by b _i, i=1,2 ..., s is assigned to from its nearest class, and upgrades the class heart i=1,2, N _iit is the sample number of the i-th class;

Step3: put q=q+1, if q>2, algorithm stops;

Step4: select the best initial center point of next class for making minimum some b _i, proceed to Step2.

3. method as claimed in claim 1 or 2, it is characterized in that, described step (1.1) specifically comprises following sub-step:

(1.1.1) obtained the 3-dimensional digital elevation model of massif by remote sensing mapping, obtain the altitude data information of true massif;

(1.1.2) structure of ANSYS geometry mountain model is formed with point, line, surface and body, point is the basis building geometric model, according to above-mentioned altitude information, the establishment of geometry massif generates closed curve by key point, plane is generated by the curve closed, then surround geometry massif by the curved surface closed, geometry massif just constitutes whole mountain model.

4. method as claimed in claim 1 or 2, it is characterized in that, described step (2.1) specifically comprises following sub-step:

(2.1.1) establish pixel grayscale in variable r representative image, use r _krepresent discrete gray levels, with P (r _k) represent probability density function, have following formula to set up:

$P_r\left(r_k\right)=\frac{n_k}{n},(0\≤r_k\≤l)$

k＝0,1,2...l-1

N in formula _kfor there is r in image _kthe pixel count of gray scale, n is pixel count sum in image, be exactly the frequency in theory of probability, l is the total number of gray level;

Known extraneous factor impact accounts for area occupied in entire image than being P%, then has following formula:

P _r(r _k)≥P

Accumulative grey level histogram successively, if aggregate-value is more than or equal to object proportion, stops cumulative, record r _kvalue, standard as a setting.

5. method as claimed in claim 1 or 2, it is characterized in that, described step (2.2) is specially:

The gray-scale value of massif thermal field Model Background carries out linear mapping correction process according to the minimum gray value of value obtained above and image; Concrete formula is as follows:

$O=\frac{I-I_l}{I_h-I_l}(O_h-O_l)+O_l$

Wherein, I is the gray-scale value of thermal field model, I _lfor the minimum brightness gray-scale value of thermal field model, I _hfor the maximum brightness gray-scale value of thermal field model, O _lfor the minimum brightness gray-scale value of true massif infrared image, O _hfor the above-mentioned r tried to achieve _k, O is the massif background model after map correction.

6. method as claimed in claim 1 or 2, it is characterized in that, described step (3.1) specifically comprises following sub-step:

(3.1.1) difference of the back and sunny side gray scale difference value is utilized to determine the pixel of boundary, it is east-west for defining boundaries according to the information of view data, more upper and lower pixel gray-scale value, if pixel grey scale difference contiguous is up and down greater than K, formula is as follows:

G(x,y)>G(x,y-1)+K

G(x,y)>G(x,y-2)+K

G(x,y)>G(x,y-3)+K

G(x,y)>G(x,y-1)+K

If above-mentioned four inequality are set up, then (x, y) is considered as the pixel near separatrix, traversal full figure obtains all pixels near separatrix;

(3.1.2) carry out cubic polynomial least square fitting linear process to the pixel near all separatrix obtained above, detailed process is as follows:

Wherein for the cubic polynomial of least square fitting, err is error objective function, and by making, err is minimum reaches optimum cubic polynomial matching, obtains final separatrix.

7. method as claimed in claim 1 or 2, it is characterized in that, described step (3.2) specifically comprises following sub-step:

After obtaining infrared image back region and sunny side region according to separatrix, the gray-scale value of following mapping policy to infrared image infrared image sunny side region is adopted to eliminate:

${F^{\′}}_{noshadow}(i,j)=A\[m_{shadow}+\frac{F_{noshadow}(i,j)-m_{noshadow}}{{\mathrm{\σ}}_{noshadow}}*{\mathrm{\σ}}_{shadow}\]$

D＝F′ _noshadow(i,j)-F _noshadow(i,j)

F' in formula _noshadow(i, j) is the sunny side area grayscale value of the non-direct irradiation of the infrared image sun, F _noshadow(i, j) is infrared image sun direct irradiation sunny side area grayscale value, m _shadowand σ _shadowaverage and the variance of infrared image sunny side area grayscale value, m _noshadowand σ _noshadowbe average and the variance of contiguous non-infrared image infrared image back area grayscale value, A is compensation intensity coefficient, and D is the energy gray-scale value of sun direct irradiation sunny side.

8. method as claimed in claim 1 or 2, it is characterized in that, described step (3.3) specifically comprises following sub-step:

Utilize the massif background heat field model emulating and obtain as the background of the infrared figure of true massif, suppose at certain a moment t ₀target location (x ₀, y ₀, z ₀) heat radiation curved surface be BT (x, y, z, t)

BT(x,y,z,t)

＝B(x,y,z,t)+T(x,y,z,t)*R _b(x,y,z,t)

+A(x,y,z,t)+δ(x,y,z,t)

This moment target background body radiation field BT (x, y, z, t), by multi-dielectric body radiation field B (x, y, z, t), objective body radiation field T (x, y, z, t) the * R that distorted by multi-dielectric body _bthe radiant quantity δ (x, y, z, t) that (x, y, z, t), water body/terrain and air contact surfaces scatter and disappear and shine upon and affect A (x, y, z, t) and jointly produce;

The impact of target background body radiation field is produced by B (x, y, z, t) and A (x, y, z, t), has following formula:

T(x,y,z,t)＝BT(x,y,z,t)-k*B(x,y,z,t)-A(x,y,z,t)

Wherein T (x, y, z, t) be the approximate thermal radiation field of target, BT (x, y, z, t) be target background body radiation field, B (x, y, z, t) be multi-dielectric body background radiation field, A (x, y, z, t) for shining upon the energy field of impact, k is the adjustability coefficients of background radiation field; Finally obtain the filtered image of true massif infrared figure background.

9. method as claimed in claim 1 or 2, it is characterized in that, described step (1.2) is specially:

Carry out free mesh, automatically generate triangle or tetrahedral grid on the whole freely, body generates tetrahedral grid automatically, manually carry out the control of smart dimensions simultaneously.

10. method as claimed in claim 1 or 2, it is characterized in that, described step (1.3) is specially:

Massif after above-mentioned stress and strain model is carried out to the setting of load boundary condition, the elementary heat of massif heat transfer and massif-air thermal convection is utilized to transmit physical basis, the parameter of heat trnasfer is set according to the pyroconductivity K of massif and the thermal convection rate Φ of massif-air, the thermo parameters method calculating massif is solved through Ansys, finally the mapping of gray-scale map and the adjustment of temperature resolution are carried out to the thermo parameters method of massif, finally obtain massif thermal field model.