CN102193093A - System and method for detecting small burning spots of forest or grassland fires by using environmental minisatellite HJ - Google Patents

Abstract

The invention relates to a system and method for carrying out automatic detection on small burning spots of forest or grassland fires by using the infrared (IRS) data of an environmental minisatellite HJ. The method comprises the following steps: radiometric correction, brightness temperature calculation, reflectivity calculation, cloud and water recognition, potential burning spot determination, absolute burning spot determination, background characteristic analysis, relative burning spot determination, burning spot confidence, and detecting burning spot output. According to the scheme provided by the invention, based on the characteristics of the environmental minisatellite HJ, the automatic detection on forest or grassland fires can be realized only by using the infrared (IRS) data, thereby enhancing the disaster prevention and reduction capacities of domestically-produced satellites in China, and promoting the industrialized application process of domestically-produced satellites.

Description

Little fire point detection system of environment moonlet HJ forest or prairie fire and method thereof

Technical field

The present invention relates to forest or prairie fire remote sensing monitoring field, relate to little fire point detection system of forest (grassland) fire and method thereof based on environment moonlet HJ data specifically, the fire point that is used for environment moonlet HJ is surveyed automatically.

Background technology

Utilize remote sensing technology monitoring forest fire, start from the twentieth century initial stage sixties, what adopted at that time is the aviation infrared eye.Along with the development of spationautics and remote sensing application technology, the experiment and the research that utilize NOAA-AVHRR, TM, defence satellite satellite data forest fire detectings such as (DMSP) have successively been carried out.Since the success of Terra satellites transmits, MODIS data being used widely in forest fire protection, and the fire in the whole world carried out daily monitoring (Justice C O, Giglio L, Kaufman Y, etal.TheMODIS fire products[J] .Remote Sensing of Environment, 2002,83:244-262.).The 4th edition (V4) fire point of present MODIS probe algorithm (Giglio, L., Descloitres, J., Justice, C.O. ， ﹠amp; Kaufman, Y.J..An enhancedcontextual fire detection algorithm for MODIS[J] .Remote Sensing ofEnvironment, 2003,87:273-282.), it is a kind of contextual algorithms based on window search and background pixel spatial analysis technology, having good adaptability, accuracy and stability in the fire monitoring in the whole world, is the classic algorithm in the fire remote sensing monitoring.Fire point that this algorithm is primarily aimed at the whole world is surveyed and is designed, thereby, there is certain limitation (Wang W T in it aspect the monitoring of zonal fire point, Qu J J, Hao X J, etal.An improved algorithm for small and cool firedetection using MODIS data:A preliminary study in the southeasternUnited States [J] .Remote Sensing ofEnvironment.2007 (108): 163-170.), show the following aspects: the one, the decision threshold for potential fire point does not have consideration of regional difference; The 2nd, the non-fire point pixel around its hypothesis fire point has similar background information; The 3rd, there is not to consider the fire point do not detect to the influence of effective background pixel etc.Therefore the V4 algorithm could be complementary with the every performance and the conflagration area characteristic of new sensor if apply to the regional area of non-global yardstick and must carry out the redesign of algorithm during new satellite sensor.

Environment and disaster monitoring forecast on September 6th, 2008 moonlet constellation A, B star (HJ-1A/1B star) succeed in sending up in Chinese Taiyuan Satellite Launch Center.Environment and disaster monitoring forecast moonlet A, B star are two optical satellites in three satellites of China's " environment and disaster monitoring forecast moonlet constellation ".The main task of this constellation be to disaster, ecological disruption, environmental pollution carry out on a large scale, the dynamic monitoring of round-the-clock, round-the-clock; development and change trend to disaster and ecologic environment is predicted; the condition of a disaster and environmental quality are carried out fast and science is assessed; improve observation, collection, transmission and the processing power of disaster and environmental information, for relief and restoration and reconstruction and environmental protection work after emergency relief, the calamity provide scientific basis.Yet the monitoring that up to the present utilizes environment moonlet HJ to carry out forest fire mainly is a classic method of utilizing visual interpretation, also is in development at the auto fire-detection method of environment moonlet HJ.

Summary of the invention

The object of the invention is to disclose little fire automatic detection system of point and the method thereof that a kind of infrared (IRS) data of utilizing environment moonlet HJ are carried out forest or prairie fire.

The present invention is achieved through the following technical solutions:

A kind of infrared (IRS) data of utilizing environment moonlet HJ are carried out the little fire automatic detection system of point or the method for forest or prairie fire, comprise with lower module or step:

A, radiant correction:

The gray-scale value (DN) of environment moonlet HJ infrared (IRS) image is converted into the radiance value with physical significance, and its physical unit is: Wm ^-2Sr ^-1

The absolute calibration coefficient g value of HJ-IRS camera wave band 1 and wave band 2 is respectively 4.2857 and 18.5579, utilizes the absolute calibration coefficient for the formula of spoke luminance picture to be with DN value image transitions:

L=DN/g formula 1

For HJ-IRS camera wave band 3 and wave band 4, utilize the absolute calibration coefficient for the formula of spoke luminance picture to be with DN value image transitions:

L=(DN-b)/g formula 2

L is spoke brightness in the formula, and g is the absolute calibration coefficient, and b is a side-play amount; The absolute calibration coefficient g value of HJ-IRS camera wave band 3 and wave band 4 is respectively 12.662 and 61.472, and side-play amount b value is respectively 11.489 and-44.598.

B, bright temperature are calculated:

Brightness temperature abbreviates bright temperature as, " equivalence " temperature parameter of general atural object is promptly described, promptly in certain wavelength band, general atural object is compared with absolute black body, when having equal radiance, with the temperature of absolute black body temperature equivalence this moment atural object, this temperature is called the brightness temperature of atural object, draw by the planck formula derivation, as shown in the formula:

$T=\frac{\mathrm{hc}}{\mathrm{\&lambda;}k\ln \left(\frac{{2\mathrm{\&pi;<figure-callout\; id="2"\; label="hc"\; filenames="HSA00000043998900041.png,HSA00000043998900042.png"\; state="\{\{state\}\}">hc</figure-callout>}}^2}{{\mathrm{L\&pi;\&lambda;}}^5+1}\right)}$ Formula 3

In the formula: T is brightness temperature (K), and h is a Planck's constant, value 6.626 * 10 ^-34(Js); K is a Boltzmann constant, value 1.3806 * 10 ^-23(JK ^-1); C is the light velocity, value 2.998 * 10 ⁸(ms ^-1); λ is wavelength (m); π=3.14159; L is radiance (Wm ^-2Sr ^-1μ m ^-1).

The environment moonlet infrared band that is used to calculate bright temperature T4 and bright temperature T11 is respectively IRS3 and IRS4.The spatial resolution of IRS3 is 150 meters, and central wavelength lambda is 3.7 μ m, and the spatial resolution of IRS43 is 300 meters, and central wavelength lambda is 11.5 μ m.

C, reflectivity calculate:

Calculate parameter with the reflectivity of environment moonlet HJ under the common meteorological condition of 6S atmospheric correction modeling, computing formula as:

$\left\{\begin{array}{c}y=x_a\&times;L_i-x_b\\ {\mathrm{\&rho;}}_i=y/(1+x_c\&times;y)\end{array}\right.$ Formula 4

ρ _iBe the reflectivity of i wave band, L _iBe the radiance of i wave band, X _a, X _b, X _c, y is respectively reflectivity and calculates parameter.

The spatial resolution of IRS1 and IRS2 wave band is 150m, is mainly used in the mask that extracts water and cloud.IRS1 wave band X _a, X _bAnd X _cValue be respectively 0.0077,0.2502 and 0.0452, IRS2 wave band X _a, X _bAnd X _cValue to divide be 0.0062,0.0564 and 0.0753 in addition.

D, cloud and water body identification:

The water body decision condition of the IRS sensor of HJ is as follows:

Water=(ρ ₁＜0.02) and (formula 5 of T4＜272K)

In the formula: Water is a water body, ρ ₁The reflectivity of expression IRS wave band 1, T4 is the bright temperature of IRS wave band 3.

The cloud decision condition of the IRS sensor of HJ is as follows:

Cloud=(T11＜265) and (Water=0) formula 6

In the formula: Cloud is a cloud, and T11 is the bright temperature of IRS wave band 4, and Water is a water body.

E, potential fire point are judged:

Potential fire point decision threshold is 325K, satisfy the following formula condition then this pixel be potential fire point pixel.

T4＞325K formula 7

F, absolute fire point are judged:

Absolute fire point decision threshold is 360K, satisfy the following formula condition then this pixel be absolute fire point pixel.

T4＞360K formula 8

G, background characteristics analysis:

To being judged as the pixel of potential fire point, adopt the context spatial statistics method of self-adapting window to come potential fire point pixel is declared knowledge one by one.When carrying out spatial statistics, relate to effective background pixel.Described effective background pixel is meant that the pixel with potential fire point is center and the pixel that satisfies following four conditions: 1) remotely-sensed data of being obtained is non-corrupt data; 2) this pixel is the land pixel; 3) this pixel is non-cloud pixel; 4) this pixel is non-background fire point pixel, and background fire point pixel is meant the pixel of T4＞325K and T4-T11＞20K.With potential fire point pixel is the center, from 5 * 5,7 * 7, searches 21 * 21 window size successively, when the quantity of effective background pixel account for whole window pixel quantity 25% the time stop search.

H, fire point judgement relatively:

After obtaining effective background pixel, adopt context spatial statistics method that fiery relatively point is judged.If this fiery relatively point satisfies a in the following formula, b, four conditions of c, d simultaneously, then, then this pixel is a fire point pixel.

$\left\{\begin{array}{c}a:(T4-T11)>\mathrm{AVG}(T4-T11)+3.5\mathrm{\&delta;}(T4-T11)\\ b:(T4-T11)>\mathrm{AVG}(T4-T11)+6K\\ c:\mathrm{<figure-callout\; id="4"\; label="T"\; filenames="HSA00000043998900042.png"\; state="\{\{state\}\}">T</figure-callout>}4>\mathrm{AVG}\left(T4\right)+3\mathrm{\&delta;}\left(T4\right)\\ d:[\mathrm{<figure-callout\; id="11"\; label="T"\; filenames="HSA00000043998900041.png"\; state="\{\{state\}\}">T</figure-callout>}11>\mathrm{AVG}\left(T11\right)+\mathrm{\&delta;}\left(T11\right)-4K)\left]\mathrm{or}\right[{\mathrm{\&delta;}}^{\&prime;}\left(T4\right)>5k]\end{array}\right.$ Formula 9

In the formula: AVG (T4-T11) is the mean value of the bright temperature difference of effective background pixel T4-T11, δ (T4-T11) is the mean absolute deviation of the bright temperature difference of effective background pixel T4-T11, AVG (T4) and AVG (T11) are respectively the mean value of effective background pixel T4 and the bright temperature of T11, δ (T4) and δ (T11) are respectively the mean absolute deviation of effective background pixel T4 and the bright temperature of T11, and δ ' is the mean absolute deviation of the bright temperature of background fire point pixel T4 (T4).

I, fiery pixel confidence:

Degree of confidence is to show that the error of sampling index and overall objective is no more than the probability assurance degree of certain limit.Under the large sample condition, degree of confidence is the function of degree of probability, and degree of probability is big more, and degree of confidence is big more.Use the bright temperature T of IRS wave band 3 ₄, the bright temperature difference T=T of IRS wave band 3 and wave band 4 ₄-T ₁₁, the water body pixel quantity Naw in 8 neighborhoods around fire point when background characteristics is analyzed, the parameters such as the quantity Nac of cloud body image unit in 8 neighborhoods around fire point when background characteristics is analyzed.And definition Z ₄And Z _{Δ T}Two variablees, as formula 10, shown in the formula 11:

${\mathrm{<figure-callout\; id="4"\; label="Z"\; filenames="HSA00000043998900042.png"\; state="\{\{state\}\}">Z</figure-callout>}}_4=\frac{T_4-\mathrm{AVG}\left(T_4\right)}{{\mathrm{\&delta;}}_4}$ Formula 10

$Z_{\mathrm{\&Delta;T}}=\frac{\mathrm{\&Delta;T}-\mathrm{AVG}\left(\mathrm{\&Delta;T}\right)}{{\mathrm{\&sigma;}}_{\mathrm{\&Delta;T}}}$ Formula 11

When carrying out confidence calculations, also use following function S (x; α, β):

$S(x;\mathrm{\&alpha;},\mathrm{\&beta;})=\left\{\begin{array}{c}0;x\&le;\mathrm{\&alpha;}\\ (x-\mathrm{\&alpha;})/(\mathrm{\&beta;}-\mathrm{\&alpha;});\mathrm{\&alpha;}<x<\mathrm{\&beta;}\\ 1;x\&GreaterEqual;\mathrm{\&beta;}\end{array}\right.$ Formula 12

The degree of confidence C of each fire point pixel is drawn by five sub-degree of confidence combination calculation, and these five sub-degree of confidence are expressed as C respectively ₁, C ₂, C ₃, C ₄, C ₅, the scope of their values be 0 (low confidence) between 1 (high confidence level), calculate by following formula respectively.

C ₁=S (T ₂₂302K, 340K) formula 13

C ₂=S (Z ₂2.5,6) and formula 14

C ₃=S (Z _{Δ T}3,6) formula 15

C ₄=1-S (Nac; 0,6) formula 16

C ₅=1-S (Naw; 0,6) formula 17

$C=\sqrt[5]{C_1{C}_2{C}_3{C}_4{\mathrm{<figure-callout\; id="5"\; label="C"\; filenames="HSA00000043998900042.png"\; state="\{\{state\}\}">C</figure-callout>}}_5}$ Formula 18

Calculate the degree of confidence of each fire point by formula 18.The value of degree of confidence is between the 0-1, and degree of confidence is low more, and the probability of actual breaking out of fire is low more; Otherwise degree of confidence is high more, and the probability of actual breaking out of fire is high more.

J, the output of detection fire point:

According to the result of detection of fire point, relevant informations such as the coordinate of fire point, bright gentle degree of confidence are surveyed in output.

Description of drawings

Fig. 1 radiant correction;

The bright temperature of Fig. 2 is calculated;

Fig. 3 reflectivity calculates;

The information extraction of Fig. 4 water body;

The information extraction of Fig. 5 cloud;

Fig. 6 background characteristics is analyzed;

The output of Fig. 7 fire point;

Fig. 8 is with the scene of a fire area error curve of zone while phase HJ-IRS and MODIS monitoring;

Fire point and image thereof that Fig. 9 Heilungkiang Zhan He area forest fire HJ-IRS surveys;

The process flow diagram of Figure 10 the method for the invention.

The present invention only just can carry out the full-automatic detection of fire point based on the HJ-IRS data.The present invention judges, judgements of absolute fire point, background characteristics analysis, judgements of relative fire point, fiery pixel confidence, surveys and fieryly put steps such as output by radiant correction, calculatings of bright temperature, reflectivity calculating, cloud and water body identification step, potential fire point, thus the fiery dot information of automatic identification extraction.Therefore, the present invention has developed little fire point detection system of environment moonlet HJ forest (grassland) fire and method thereof according to the parameter characteristic of infrared (IRS) sensor of environment moonlet HJ and the district characteristic of China forest or prairie fire.

Embodiment:

The invention process case is chosen as the IRS image of the forest fire environment moonlet HJ in the Zhan He area, China Dark Longjiang that took place April 29 in 2009.

1. system flow

A, radiant correction:

(unit is: Wm to be used for that the gray-scale value (DN) of environment moonlet HJ infrared (IRS) image is transformed the radiance value of proofreading and correct to having physical significance ^-2Sr ^-1)。

B, bright temperature are calculated:

Be used for the IRS wave band 3 of computing environment moonlet HJ and the bright temperature T4 and the T11 of wave band 4.

C, reflectivity calculate:

Be used for the infrared IRS wave band 1 of computing environment moonlet HJ and the reflectivity ρ of wave band 2 ₁And ρ ₂

D, cloud and water body identification:

Be used to carry out the identification of cloud and water body, generate the mask of cloud and water body.

The water body decision condition of the IRS sensor of HJ is as follows:

Water=(ρ ₁＜0.02) and (formula of T4＜272K)

In the formula: Water is a water body, ρ ₁The reflectivity of expression IRS wave band 1, T4 is the bright temperature of IRS wave band 3.

The cloud decision condition of the IRS sensor of HJ is as follows:

Cloud=(T11＜265) and (Water=0) formula

In the formula: Cloud is a cloud, and T11 is the bright temperature of IRS wave band 4, and Water is a water body.

E, potential fire point are judged:

Be used to judge whether pixel belongs to potential fiery point.

Potential fire point decision threshold is 325K, satisfy the following formula condition then this pixel be potential fire point pixel.

T4＞325K formula 25

F, absolute fire point are judged:

Be used to judge whether pixel belongs to absolute fiery point.

Absolute fire point decision threshold is 360K, satisfy the following formula condition then this pixel be absolute fire point pixel.

T4＞360K formula 26

G, background characteristics analysis:

Be used for the surrounding environment of potential fire point pixel is carried out the spatial statistics analysis.

To being judged as the pixel of potential fire point, adopt the context spatial statistics method of self-adapting window to come potential fire point pixel is declared knowledge one by one.Effectively the background pixel is meant that the pixel with potential fire point is center and the pixel that satisfies following four conditions: 1) remotely-sensed data of being obtained is non-corrupt data; 2) this pixel is the land pixel; 3) this pixel is non-cloud pixel; 4) this pixel is non-background fire point pixel, and background fire point pixel is meant the pixel of T4＞325K and T4-T11＞20K.With potential fire point pixel is the center, from 5 * 5,7 * 7, searches 21 * 21 window size successively, when the quantity of effective background pixel account for whole window pixel quantity 25% the time stop search.

H, fire point judgement relatively:

Carry out the judgement of fire point relatively with crossing.

After obtaining effective background pixel, adopt context spatial statistics method that fiery relatively point is judged.If this fiery relatively point satisfies a in the following formula, b, four conditions of c, d simultaneously, then, then this pixel is a fire point pixel.

$\left\{\begin{array}{c}a:(T4-T11)>\mathrm{AVG}(T4-T11)+3.5\mathrm{\&delta;}(T4-T11)\\ b:(T4-T11)>\mathrm{AVG}(T4-T11)+6K\\ c:\mathrm{<figure-callout\; id="4"\; label="T"\; filenames="HSA00000043998900042.png"\; state="\{\{state\}\}">T</figure-callout>}4>\mathrm{AVG}\left(T4\right)+3\mathrm{\&delta;}\left(T4\right)\\ d:[\mathrm{<figure-callout\; id="11"\; label="T"\; filenames="HSA00000043998900041.png"\; state="\{\{state\}\}">T</figure-callout>}11>\mathrm{AVG}\left(T11\right)+\mathrm{\&delta;}\left(T11\right)-4K)\left]\mathrm{or}\right[{\mathrm{\&delta;}}^{\&prime;}\left(T4\right)>5k]\end{array}\right.$ Formula 27

In the formula: AVG (T4-T11) is the mean value of the bright temperature difference of effective background pixel T4-T11, δ (T4-T11) is the mean absolute deviation of the bright temperature difference of effective background pixel T4-T11, AVG (T4) and AVG (T11) are respectively the mean value of effective background pixel T4 and the bright temperature of T11, δ (T4) and δ (T11) are respectively the mean absolute deviation of effective background pixel T4 and the bright temperature of T11, and δ ' is the mean absolute deviation of the bright temperature of background fire point pixel T4 (T4).

I, fiery pixel confidence:

Be used to calculate the degree of confidence of fire point.

J, the output of detection fire point:

Be used for the result of detection of fire point is exported with the form of text.

2. the superiority of interpretation of result and this method

Utilize the forest fire in the Zhan He area, China Dark Longjiang that takes place the 27-5 month 2 in April, 2009, choose HJ-IRS and the MODIS data of several different periods with the forest fire that occurred in California, USA May 8 in 2009, select a plurality of test zones (partly to carry out the zone cutting by image overlap, result to HJ-IRS and the detection of MODIS fire point compares analysis, obtain fire point situation with zone while phase HJ-IRS and MODIS detection, and utilize the burnt area data in each scene of a fire of field observation and statistics that the monitoring result of above-mentioned two kinds of methods is compared analysis, statistics sees Table 1 and Fig. 8.

Table 1

Table 1 is a scene of a fire area error situations of surveying quantity, scene of a fire remote sensing monitoring area, scene of a fire real area and two kinds of remote sensing means monitorings with the fire point of zone while phase HJ-IRS and MODIS.

The result of fire point that HJ-IRS surveys and the stack of IRS image as shown in Figure 9.Scene of a fire area error mean value with zone while phase HJ-IRS and MODIS monitoring is respectively 4.56% and 18.24%, as seen the precision aspect of the scene of a fire area of surveying than MODIS with the scene of a fire area of HJ-IRS monitoring will exceed 13.68%, has higher forest (grassland) fire monitoring precision.Compare with the visual determination methods of traditional expert, then have high efficiency based on the automatic detection method of HJ-IRS forest (grassland) fire point, the personnel that need not are on duty, report to the police automatically and successfully wait multiple advantages.This method will help improving application level and the professional ability of environment moonlet HJ in China forest (grassland) fire remote sensing monitoring field.

Claims (2)

1. an infrared data that utilizes environment moonlet HJ carries out the little fire automatic detection system of point or the method for forest or prairie fire, it is characterized in that comprising with lower module or step:

A, radiant correction:

The gray-scale value (DN) of environment moonlet HJ infrared (IRS) image is converted into the radiance value with physical significance;

B, bright temperature are calculated:

The environment moonlet infrared band that is used to calculate bright temperature T4 and bright temperature T11 is respectively IRS3 and IRS4; The spatial resolution of IRS3 is 150 meters, and central wavelength lambda is 3.7 μ m, and the spatial resolution of IRS43 is 300 meters, and central wavelength lambda is 11.5 μ m;

C, reflectivity calculate:

Reflectivity with environment moonlet HJ under the common meteorological condition of 6S atmospheric correction modeling calculates parameter;

D, cloud and water body identification:

The water body decision condition of the IRS sensor of HJ is as follows:

Water＝(ρ ₁＜0.02)and(T4＜272K)

In the formula: Water is a water body, ρ ₁The reflectivity of expression IRS wave band 1, T4 is the bright temperature of IRS wave band 3;

The cloud decision condition of the IRS sensor of HJ is as follows:

Cloud＝(T11＜265)and(Water＝0)

In the formula: Cloud is a cloud, and T11 is the bright temperature of IRS wave band 4, and Water is a water body;

E, potential fire point are judged:

Potential fire point decision threshold is 325K;

F, absolute fire point are judged:

Absolute fire point decision threshold is 360K;

G, background characteristics analysis:

To being judged as the pixel of potential fire point, adopt the context spatial statistics method of self-adapting window to come potential fire point pixel is declared knowledge one by one;

H, fire point judgement relatively:

After obtaining effective background pixel, adopt context spatial statistics method that fiery relatively point is judged;

I, fiery pixel confidence:

Degree of confidence is meant that the error of sampling index and overall objective is no more than the probability assurance degree of certain limit; The value of degree of confidence is between the 0-1, and degree of confidence is low more, and the probability of actual breaking out of fire is low more; Otherwise degree of confidence is high more, and the probability of actual breaking out of fire is high more;

J, the output of detection fire point:

According to the result of detection of fire point, coordinate, the bright gentle degree of confidence relevant information of fire point surveyed in output.

2. a kind of infrared data that utilizes environment moonlet HJ as claimed in claim 1 carries out the little fire automatic detection system of point or the method for forest or prairie fire, it is characterized in that comprising with lower module or step:

A, radiant correction:

The gray-scale value (DN) of environment moonlet HJ infrared (IRS) image is converted into the radiance value with physical significance, and its physical unit is: Wm ^-2Sr ^-1

The absolute calibration coefficient g value of HJ-IRS camera wave band 1 and wave band 2 is respectively 4.2857 and 18.5579, utilizes the absolute calibration coefficient for the formula of spoke luminance picture to be with DN value image transitions:

L=DN/g formula 1

For HJ-IRS camera wave band 3 and wave band 4, utilize the absolute calibration coefficient for the formula of spoke luminance picture to be with DN value image transitions:

L=(DN-b)/g formula 2

L is spoke brightness in the formula, and g is the absolute calibration coefficient, and b is a side-play amount; The absolute calibration coefficient g value of HJ-IRS camera wave band 3 and wave band 4 is respectively 12.662 and 61.472, and side-play amount b value is respectively 11.489 and-44.598;

B, bright temperature are calculated:

Brightness temperature is derived by planck formula and is drawn, as shown in the formula:

$T=\frac{\mathrm{hc}}{\mathrm{\&lambda;}k\ln \left(\frac{2\mathrm{\&pi;h}{c}^2}{\mathrm{L\&pi;}{\mathrm{\&lambda;}}^5+1}\right)}$ Formula 3

In the formula: T is brightness temperature (K), and h is a Planck's constant, value 6.626 * 10 ^-34(Js); K is a Boltzmann constant, value 1.3806 * 10 ^-23(JK ^-1); C is the light velocity, value 2.998 * 10 ⁸(ms ^-1); λ is wavelength (m); π=3.14159; L is radiance (Wm ^-2Sr ^-1μ m ^-1);

The environment moonlet infrared band that is used to calculate bright temperature T4 and bright temperature T11 is respectively IRS3 and IRS4; The spatial resolution of IRS3 is 150 meters, and central wavelength lambda is 3.7 μ m, and the spatial resolution of IRS43 is 300 meters, and central wavelength lambda is 11.5 μ m;

C, reflectivity calculate:

Calculate parameter with the reflectivity of environment moonlet HJ under the common meteorological condition of 6S atmospheric correction modeling, computing formula as:

$\left\{\begin{array}{c}y=x_a\&times;L_i-x_b\\ {\mathrm{\&rho;}}_i=y/(1+x_c\&times;y)\end{array}\right.$ Formula 4

ρ _iBe the reflectivity of i wave band, L _iBe the radiance of i wave band, X _a, X _b, X _c, y is respectively reflectivity and calculates parameter;

The spatial resolution of IRS1 and IRS2 wave band is 150m, is mainly used in the mask that extracts water and cloud; IRS1 wave band X _a, X _bAnd X _cValue be respectively 0.0077,0.2502 and 0.0452, IRS2 wave band X _a, X _bAnd X _cValue be respectively 0.0062,0.0564 and 0.0753;

D, cloud and water body identification:

The water body decision condition of the IRS sensor of HJ is as follows:

Water=(ρ ₁＜0.02) and (formula 5 of T4＜272K)

In the formula: Water is a water body, ρ ₁The reflectivity of expression IRS wave band 1, T4 is the bright temperature of IRS wave band 3;

The cloud decision condition of the IRS sensor of HJ is as follows:

Cloud=(T11＜265) and (Water=0) formula 6

In the formula: Cloud is a cloud, and T11 is the bright temperature of IRS wave band 4, and Water is a water body;

E, potential fire point are judged:

Potential fire point decision threshold is 325K, satisfy the following formula condition then this pixel be potential fire point pixel;

T4＞325K formula 7

F, absolute fire point are judged:

Absolute fire point decision threshold is 360K, satisfy the following formula condition then this pixel be absolute fire point pixel;

T4＞360K formula 8

G, background characteristics analysis:

To being judged as the pixel of potential fire point, adopt the context spatial statistics method of self-adapting window to come potential fire point pixel is declared knowledge one by one; When carrying out spatial statistics, relate to effective background pixel; Described effective background pixel is meant that the pixel with potential fire point is center and the pixel that satisfies following four conditions: the remotely-sensed data of being obtained is non-corrupt data; This pixel is the land pixel; This pixel is non-cloud pixel; This pixel is non-background fire point pixel, and background fire point pixel is meant the pixel of T4＞325K and T4-T11＞20K; With potential fire point pixel is the center, from 5 * 5,7 * 7, searches 21 * 21 window size successively, when the quantity of effective background pixel account for whole window pixel quantity 25% the time stop search;

H, fire point judgement relatively:

After obtaining effective background pixel, adopt context spatial statistics method that fiery relatively point is judged; If this fiery relatively point satisfies a in the following formula, b, four conditions of c, d simultaneously, then, then this pixel is a fire point pixel;

$\left\{\begin{array}{c}a:(T4-T11)>\mathrm{AVG}(T4-T11)+3.5\mathrm{\&delta;}(T4-T11)\\ b:(T4-T11)>\mathrm{AVG}(T4-T11)+6K\\ c:T4>\mathrm{AVG}\left(T4\right)+3\mathrm{\&delta;}\left(T4\right)\\ d:[T11>\mathrm{AVG}\left(T11\right)+\mathrm{\&delta;}\left(T11\right)-4K)]\mathrm{or}[{\mathrm{\&delta;}}^{\&prime;}\left(T4\right)>5k]\end{array}\right.$ Formula 9

In the formula: AVG (T4-T11) is the mean value of the bright temperature difference of effective background pixel T4-T11, δ (T4-T11) is the mean absolute deviation of the bright temperature difference of effective background pixel T4-T11, AVG (T4) and AVG (T11) are respectively the mean value of effective background pixel T4 and the bright temperature of T11, δ (T4) and δ (T11) are respectively the mean absolute deviation of effective background pixel T4 and the bright temperature of T11, and δ ' is the mean absolute deviation of the bright temperature of background fire point pixel T4 (T4);

I, fiery pixel confidence:

Degree of confidence is to show that the error of sampling index and overall objective is no more than the probability assurance degree of certain limit; Under the large sample condition, degree of confidence is the function of degree of probability, and degree of probability is big more, and degree of confidence is big more; Use the bright temperature T of IRS wave band 3 ₄, the bright temperature difference T=T of IRS wave band 3 and wave band 4 ₄-T ₁₁, the water body pixel quantity Naw in 8 neighborhoods around fire point when background characteristics is analyzed, the parameters such as the quantity Nac of cloud body image unit in 8 neighborhoods around fire point when background characteristics is analyzed; And definition Z ₁And Z _{Δ T}Two variablees, as formula 10, shown in the formula 11:

$Z_4=\frac{T_4-\mathrm{AVG}\left(T_4\right)}{{\mathrm{\&delta;}}_4}$ Formula 10

$Z_{\mathrm{\&Delta;T}}=\frac{\mathrm{\&Delta;T}-\mathrm{AVG}\left(\mathrm{\&Delta;T}\right)}{{\mathrm{\&sigma;}}_{\mathrm{\&Delta;T}}}$ Formula 11

When carrying out confidence calculations, also use following function S (x; α, β):

$S(x;\mathrm{\&alpha;},\mathrm{\&beta;})=\left\{\begin{array}{c}0;x\&le;\mathrm{\&alpha;}\\ (x-\mathrm{\&alpha;})\\ 1;x\&GreaterEqual;\mathrm{\&beta;}\end{array},/(\mathrm{\&beta;}-\mathrm{\&alpha;});\mathrm{\&alpha;}<x<\mathrm{\&beta;}\right.$ Formula 12

The degree of confidence C of each fire point pixel is drawn by five sub-degree of confidence combination calculation, and these five sub-degree of confidence are expressed as C respectively ₁, C ₂, C ₃, C ₄, C ₅, the scope of their values be 0 (low confidence) between 1 (high confidence level), calculate by following formula respectively;

C ₁=S (T ₂₂302K, 340K) formula 13

C ₂=S (Z ₂2.5,6) and formula 14

C ₃=S (Z _{Δ T}3,6) formula 15

C ₄=1-S (Nac; 0,6) formula 16

C ₅=1-S (Naw; 0,6) formula 17

$C=\sqrt[5]{C_1{C}_2{C}_3{C}_4{C}_5}$ Formula 18

Calculate the degree of confidence of each fire point by formula 18; The value of degree of confidence is between the 0-1, and degree of confidence is low more, and the probability of actual breaking out of fire is low more; Otherwise degree of confidence is high more, and the probability of actual breaking out of fire is high more;

J, the output of detection fire point:

According to the result of detection of fire point, coordinate, the bright gentle degree of confidence relevant information of fire point surveyed in output.

Images