CN105627997A - Multi-angle remote sensing water depth decision fusion inversion method - Google Patents

Abstract

The invention relates to a multi-angle remote sensing water depth decision fusion inversion method which comprises the following steps: step 1, multi-spectral remote sensing image pre-processing so as to obtain sea surface reflectance; step 2, field actually-measured water depth value acquisition and processing; step 3, single-angle water depth inversion and water depth section identification; step 4, multi-angle water depth inversion fusion; step 5, water depth inversion precision verification, wherein after the water depth inversion precision verification is finished, a final water depth value is taken as remote sensing image actual water depth value output data. Compared with an existing inversion method, the multi-angle remote sensing water depth decision fusion inversion method has the benefit that by the utilization of multiple imaging capabilities of multi-angle remote sensing under different sea conditions, different substrates and different water qualities, rich data information can be brought to water depth remote sensing inversion, so that the inversion precision is improved, and the multi-angle remote sensing water depth decision fusion inversion method is particularly suitable for ocean water depth measurement in shallow water areas under complicated situations.

Description

Multiple-angle thinking depth of water Decision fusion inversion method

Technical field

The present invention relates to a kind of ocean water deep investigation method, belong to space remote sensing technical field, particularly relate to a kind of multiple-angle thinking depth of water Decision fusion inversion method.

Background technology

Ocean depth of water DATA REASONING is guarantee ship's navigation, carries out port and pier and the necessary basis data of ocean engineering construction, formulation seashore and island Correlative plan. Compared with depth of water in-site measurement means, remote sensing technology has the advantage that covering is wide, the cycle is short, expense is low, spatial resolution is high. Since 20 century 70s, having carried out the research of various passive remote sensing Depth extraction model both at home and abroad, conventional visible ray Depth extraction model mainly includes analytical model, half analysis semiempirical model and statistical model. Utilize different model, in recent years in river, lake, reservoir, the water-depth measurement field such as island and littoral zone periphery carried out inverting application.

Depth of water visual remote sensing inverting is the favourable solution obtaining the shallow sea complexity landform depth of water, it is particularly possible to inverting obtain ship cannot near and be difficult to enter the water depth information in region. But owing to model is difficult to take into account physical mechanism and parametrization, the limited space that therefore existing visible ray RS Fathoming inverse model precision improves again.

For identical area, acquisition time fairly close again many scapes multi-angle image often comprises the information such as the stereochemical structure of a large amount of Target scalar. Data of multiple angles has had more application in Reflectivity Model and the extraction of product, image classification and biomass estimation etc., but not yet has the application in RS Fathoming inverting of the multi-angle image. Utilize multiple-angle thinking to many-sided imaging capability under different sea situations, different substrate and different quality situation, abundant data message can be brought for RS Fathoming inverting; Meanwhile, adopting Decision fusion also to be able to make full use of existing remote sensing image resource and information, improve the precision of Depth extraction, therefore, the shallow water depth multiple-angle thinking image inversion method based on Decision fusion has important using value.

Chinese patent (application number 201310188829.4, Shen Qing Publication day CN104181515A) discloses " a kind of shallow water depth inversion method based on blue-yellow wave band high-spectral data ". it is mainly used in solving to utilize the model that optical remote sensing means are cleaned water body Depth extraction mostly to set up for multispectral data, such algorithm is subject to multispectral data wide waveband, the restriction that spectral information is few, this invention is according to water body optical attenuation mechanism, propose one based on high-spectral data and utilize the new method of blue-yellow wave band (450-610 nanometer) high-spectral data inverting cleaning water body shallow water depth, the method can accurately extract 30 meters within shallow water depth distributed intelligence, and for a kind of remote sensor, have only to carry out an algorithm coefficient demarcate, algorithm universality be improved significantly. but the method adopts single angle image of the remote sensor acquisition in single source as detection data source, available remote sensing image spectral information is limited in scope, it is unfavorable for the accuracy improving shallow water depth inverting for water-depth measurement, especially in complex situations that the Effect on Detecting of the neritic province domain depth of water is not enough.

Summary of the invention

The invention provides a kind of multiple-angle thinking depth of water Decision fusion inversion method, during for solving the single angle image only using single source remote sensor in prior art as data source, exist because imaging angle is not good, Water Depth Information is subject to the problem that the impact of a large amount of seas table reflection, water-depth measurement precision and accuracy are poor.

Multiple-angle thinking depth of water Decision fusion inversion method, comprises the following steps:

The first step: multi-spectrum remote sensing image is carried out pretreatment, obtains sea table reflectance;

Described pretreatment includes radiance conversion, atmospheric correction and solar flare are removed;

Second step: field measurement water depth value obtains and processes;

Obtain the bathymetric data of test block and corresponding latitude and longitude coordinates, the tidal height value measuring the moment is confirmed by tide table, bathymetric data correction is obtained the depth of water of theoretical depth datum level, further according to the acquisition moment of multi-spectrum remote sensing image, the bathymetric data of theoretical depth datum level is carried out the tidal correction of the instantaneous depth of water to obtain the instantaneous depth of water;

3rd step: single angle Depth extraction and depth of water segment identification;

According to the depth of water control point place depth of water and the relation between corresponding image picture element reflectance value, adopt multiband model to carry out statistical regression to go out, export the input that the parameter of this scape image Depth extraction merges as multi-angle Depth extraction, and multiband model is carried out parameter calibration, multiband model formation is as follows

$Z=A_0+\underset{i=1}{\overset{n}{\Σ}}{A}_i{X}_i---\left(1\right)$

X_i=Ln (��_i-��_si)(2)

Wherein, Z is the depth of water, and n is the wave band number participating in inverting, A₀And A_iFor undetermined coefficient, ��_iIt is the i-th wave band reflectivity data, ��_siIt it is the reflectance at this wave band deep water place;

Depth of water control point being divided into multiple depth of water section as input, export average relative error and the credibility of each depth of water section, another fusion as multi-angle Depth extraction inputs, i.e. fusion parameters,

${\mathrm{\δ}}_k=\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}\frac{|z_i-{z_i}^{\′}|}{z_i}---\left(3\right)$

$L_k=0.5+{\left\{\underset{i=1}{\overset{n}{\Σ}}{w}_i{(L_k^i-0.5)}^{\mathrm{\α}}\right\}}^{\frac{1}{\mathrm{\α}}}---\left(4\right)$

Wherein, k represents depth of water section, and n is depth of water checkpoint number, in formula 3, and ��_kIt is the average relative error of kth depth of water section, z_iIt is the measured value of i-th depth of water checkpoint, z_i' for its inverting value, in formula 4, L_kRepresent the credibility of kth depth of water section, w_iRepresent the relative importance of each information source, and meetIts formula is Representing the Kappa coefficient at this i scape image control point, n represents the number participating in inverting fusion evaluation, and �� is odd number,

Utilize fusion parameters and whole scape remote sensing image, calculate and obtain single angle Depth extraction result, and by its tidal correction to theoretical depth datum level, obtain depth of water segment identification image afterwards;

4th step: multi-angle Depth extraction merges;

The input single angle Depth extraction result, depth of water segment identification image and fusion parameters merged as multi-angle Depth extraction, carries out fusion by pixel, specifically includes:

(1) judging the depth of water segment identification that pixel is corresponding, next step computing of carrying out that pixel number in same depth of water section and this depth of water section is maximum, otherwise, according to (3rd) stepping row operation;

(2) judging the water depth value of pixel, if there being the pixel value of 2 or more than 2 identical, composing this value is final water depth value, if it is different, carry out next step computing;

(3) judging whether pixel is singular value in 5 �� 5 neighborhoods, selecting nonsingular value point assignment is final water depth value, if not singular point number has 2 or more than 2, carrying out next step computing, if being singular point, carrying out according to (5th) step;

(4) pixel average relative error in current depth of water section is compared, with pixel value corresponding to minimum error for final water depth value;

(5) intermediate value of pixel 5 �� 5 neighborhood is calculated, and with this neighborhood intermediate value for new pixel value, it is judged that its depth of water section belongs to, if identical, carries out according to (4th) step, otherwise carries out next step computing;

(6) the credibility size of pixel value place depth of water section is compared, with the big pixel value of credibility for final water depth value;

5th step: Depth extraction precision test;

Described precision test is the comparison of multi-angle inversion result after utilizing checkpoint to carry out the front single angle inversion result of fusion and merge, and after Depth extraction precision test completes, as the actual water depth value of remote sensing images, final water depth value is exported data.

Beneficial effects of the present invention:

The present invention is compared with existing inversion method, the present invention utilizes multiple-angle thinking to many-sided imaging capability under different sea situations, different substrate and different quality situation, abundant data message can be brought for RS Fathoming inverting, improve inversion accuracy, the marine sounding being particularly suited under complex situations shallow water area.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention;

Fig. 2 a is the forward sight image of stereogram selected by the embodiment of the present invention;

Fig. 2 b is the backsight image of stereogram selected by the embodiment of the present invention;

Fig. 3 is the Decision fusion flow chart of the embodiment of the present invention;

Fig. 4 a is embodiment of the present invention inversion result scatterplot;

Fig. 4 b is WorldView-2 forward sight image Depth extraction result scatterplot;

Fig. 4 c is WorldView-2 backsight image Depth extraction result scatterplot;

Fig. 5 is the result images of the embodiment of the present invention;

Detailed description of the invention

For making the purpose of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is a part of embodiment of the present invention, rather than whole embodiments.

In conjunction with accompanying drawing 1, the specific embodiment of the invention being described in further detail, multiple-angle thinking depth of water Decision fusion inversion method specifically includes following steps:

The first step: Multi-spectral Remote Sensing Data pretreatment:

First have to the multi-spectrum remote sensing image to participating in Depth extraction fusion and carry out pretreatment, remove including spoke brightness transition, atmospheric correction and solar flare, spoke brightness transition is the process that image DN value is converted into spoke brightness, the spoke brightness transition formula that different remotely-sensed data products is corresponding is different, is generally following two:

$L=\frac{DN}{Gain}+Bias---\left(5\right)$

$L=\frac{DN*absCalFactor}{effectiveBandwidth}---\left(6\right)$

Parameters corresponding in formula all can obtain in the meta data file of image, after obtaining multispectral spoke brightness image, adopts FLAASH or the dark atmospheric correction method such as pixel, 6S to carry out atmospheric correction, obtains sea table reflectivity data; For removing the interference that sea table solar flare and floating thing etc. bring, intermediate value or the method such as average, small echo is then adopted to carry out solar flare removal.

Second: field measurement water depth value obtains and processes:

Obtain the bathymetric data of test block and corresponding latitude and longitude coordinates, the tidal height value measuring the moment is confirmed by tide table, bathymetric data correction is obtained the depth of water of theoretical depth datum level, further according to the acquisition moment of multi-spectrum remote sensing image, the bathymetric data of theoretical depth datum level is carried out the tidal correction of the instantaneous depth of water to obtain the instantaneous depth of water;

3rd step: single angle Depth extraction and depth of water segment identification:

According to the depth of water control point place depth of water and the relation between corresponding image picture element reflectance value, adopt multiband model to carry out statistical regression to go out, export the input that the parameter of this scape image Depth extraction merges as multi-angle Depth extraction, and multiband model is carried out parameter calibration, multiband model formation is as follows

$Z=A_0+\underset{i=1}{\overset{n}{\Σ}}{A}_i{X}_i---\left(1\right)$

X_i=Ln (��_i-��_si)(2)

Wherein, Z is the depth of water, and n is the wave band number participating in inverting, A₀And A_iFor undetermined coefficient, ��_iIt is the i-th wave band reflectivity data, ��_siIt it is the reflectance at this wave band deep water place;

Depth of water control point being divided into multiple depth of water section as input, export average relative error and the credibility of each depth of water section, another fusion as multi-angle Depth extraction inputs, i.e. fusion parameters,

${\mathrm{\δ}}_k=\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}\frac{|z_i-{z_i}^{\′}|}{z_i}---\left(3\right)$

$L_k=0.5+{\left\{\underset{i=1}{\overset{n}{\Σ}}{w}_i{(L_k^i-0.5)}^{\mathrm{\α}}\right\}}^{\frac{1}{\mathrm{\α}}}---\left(4\right)$

Wherein, k represents depth of water section, and formula 3 is average relative error model formation, z_iIt is the measured value of i-th depth of water checkpoint, z_i' for its inverting value, n is depth of water checkpoint number, and formula 4 is credibility model formula, L_kRepresenting the credibility of kth depth of water section, wj represents the relative importance of each information source, and meetsN represents the number participating in inverting fusion evaluation, and �� is odd number. As adopted 2 scape image, i.e. n=2, its nicety of grading respectively a and b, thenA and the b Kappa coefficient calculations at control point herein.

Utilize fusion parameters and whole scape remote sensing image, calculating obtains single angle Depth extraction result, and by its tidal correction to theoretical depth datum level, theoretical depth datum level is equivalent to one and starts at face, on this datum level, the depth of water does not include the image that tide brings, and obtains depth of water segment identification image afterwards. As the depth of water of 0-20m depth bounds carried out segmentation with 2m, 5m, 10m for interval, it is divided into into 4 sections of (i.e. k values [1 of formula 3 and formula 4,4]), then depth of water segment identification image is 1 the depth of water scope meaning this point (0,2] in m, the depth of water scope meaning this point of 2 (2,5] in m, and so on.

4th step: multi-angle Depth extraction merges:

The input single angle Depth extraction result, depth of water segment identification image, cloud mask and fusion parameters merged as multi-angle Depth extraction, carries out fusion by pixel:

(1) depth of water segment identification that pixel is corresponding is judged, in same depth of water section and belong to next step computing of carrying out that the pixel number of this depth of water section is maximum together, otherwise, according to the 3rd stepping row operation;

(2) judging the water depth value of pixel, if there being the pixel value of 2 or more than 2 identical, composing this value is final water depth value, if it is different, carry out next step computing;

(3) judging whether pixel is singular value in 5 �� 5 neighborhoods, selecting nonsingular value point assignment is final water depth value, if not singular point number has 2 or more than 2, carrying out next step computing, if being singular point, carrying out according to the 5th step;

(4) pixel average relative error in current depth of water section is compared, with pixel value corresponding to minimum error for final water depth value;

(5) intermediate value of pixel 5 �� 5 neighborhood is calculated, and with this neighborhood intermediate value for new pixel value, it is judged that its depth of water section belongs to, if identical, carries out according to the 4th step, otherwise carries out next step computing;

(6) the credibility size of pixel value place depth of water section is compared, with the big pixel value of credibility for final water depth value.

As shown in Fig. 2 a, 2b, have chosen 2 scape WorldView-2 images in the present embodiment, be taken at 03:35:27 on October 25 (UTC) in 2012, concrete Decision fusion process includes:

I and j represents forward sight and backsight image, C_iWith C_jThe respectively class label of a certain pixel, Z on 2 scape depth of water mark images_iWith Z_jRepresenting current pixel water depth value respectively, max (), min (), med () represent the depth of water maximum in 5 �� 5 neighborhoods centered by current pixel, minima and intermediate value respectively,WithRepresent the current pixel average relative error in corresponding depth section respectively,WithRepresent the fuzzy membership of current pixel correspondence place depth of water section respectively,WithRepresent the depth of water section at current pixel neighborhood intermediate value place respectively, in like manner,WithIt is the average relative error degree of membership with this depth of water section of i pixel neighborhood intermediate value place depth of water section respectively. Z_sFor the water depth value after Decision fusion.

As it is shown on figure 3, mainly consider 2 kinds of processes result during Decision fusion, namely first is that the depth of water section that inverting obtains is identical, and second is that the 2 scape image inverting depth of waters belong to different water depth section.

The first, C is worked as_i=C_jIf, Z_i=Z_j, then Z_s=Z_i=Z_j;

Work as C_i=C_j, and Z_i��Z_jIf, min (Z_i)<Z_i<max(Z_i), Z_j<min(Z_j) or Z_j>max(Z_j), then Z_s=Z_iOtherwise, if Z_i<min(Z_i) or Z_i>max(Z_i), min (Z_j)<Z_j<max(Z_j), then Z_s=Z_j;

Work as C_i=C_j, and Z_i��Z_j, min (Z_i)<Z_i<max(Z_i), min (Z_j)<Z_j<max(Z_j), ifThen Z_s=Z_iOtherwise, then Z_s=Z_j;

Work as C_i=C_j, and Z_i��Z_j, Z_i<min(Z_i) or Z_i>max(Z_i), Z_j<min(Z_j) or Z_j>max(Z_j), try to achieve med (Z_i) and med (Z_j), ifAndThen Z_s=med (Z_i), otherwise,Then Z_s=med (Z_j);

Work as C_i=C_j, and Z_i��Z_j, Z_i<min(Z_i) or Z_i>max(Z_i), Z_j<min(Z_j) or Z_j>max(Z_j), try to achieve med (Z_i) and med (Z_j), ifAndThen Z_s=med (Z_i), otherwise, then Z_s=med (Z_j);

The second, C is worked as_i��C_jIf, $max(L_{C_i},L_{C_j})\&GreaterEqual;0.2,$ When $max(L_{C_i},L_{C_j})=L_{C_i}$ And $P_{C_i}<{P^{\&prime;}}_{C_i},$ Then Z_s=Z_i, otherwise, whenAndThen Z_s=Z_j;

Work as C_i��C_jIf, $max(L_{C_i},L_{C_j})\&GreaterEqual;0.2,$ When $max(L_{C_i},L_{C_j})=L_{C_j}$ And $P_{C_i}<{P^{\&prime;}}_{C_i},$ Then Z_s=Z_j, otherwise, whenAndThen Z_s=Z_i;

Work as C_i��C_jIf,WhenThen Z_s=Z_iOtherwise, then Z_s=Z_j��

In above-mentioned expression formula, min (Z_i)<Z_i<max(Z_i) represent water depth value nonsingular value in the neighborhood of 5 �� 5 windows of current pixel,Represent the average relative error of the depth of water section belonging to current pixel 5 �� 5 neighborhood intermediate value of i image and this depth of water section respectively.

5th step: Depth extraction precision test:

The precision test of multi-angle inversion result after utilizing checkpoint to carry out the front single angle inversion result of fusion and merge, calculate average relative error that is overall and that divide different water depth section and mean absolute error, thus the precision of multi-angle Depth extraction Fusion Model is verified.

(1) depth of water multiple-angle thinking inverting fusion parameters and Fusion Model implementation status

The present embodiment is chosen the WorldView-2 stereogram on October 25th, 2012 and is carried out depth of water multi-angle inverting fusion experiment. Table 1 is illustrated the inverted parameters obtained by three wave band log-linear model Depth extraction blue, green, red and the segmental averaging relative error at control point place. Depth of water multi-angle merges the parameter used in inverting and includes Kappa coefficient and the segmental averaging relative error of WorldView-2 stereogram Depth extraction, and each inverted parameters and fusion parameters are referring to table 1.

Table 1 depth of water list angle remote-sensing inversion parameter and three-dimensional Decision fusion parameter

Relatively Kappa coefficient can find that WorldView-2 forward sight image is substantially better than seeing image picture thereafter in overall segmentation precision; Observing segmental averaging relative error, forward sight image, except interior with nearly 16 percentage points relatively except backsight image difference at the 1st section, is all substantially better than backsight image at 2,3,4 sections.

WorldView-2 stereogram totally 1000 row, 1250 row. Because relating to the neighborhood operation of 5 �� 5 in merging, starting to the 998th row 1248 row to terminate from the 3rd row the 3rd row, namely having 1241016 pixels can be assigned to new water depth value through Decision fusion. In depth of water solid inverting fusion experiment, it is judged to that the pixel number of same depth of water section has 877398 for the first time, accounting is 70.7%, what wherein execution number of times was maximum is the 2nd and the 3rd rule, the pixel having 88.3% determines final water depth value by this rule, and what perform least number of times is that the 1st rule only has the forward sight Depth extraction image of 1436 pixels and backsight Depth extraction image all to obtain identical water depth value. Having 363618 pixels by 2 scape scope interpretations for not in same depth of water section, this wherein has the pixel of 81.1% finally to determine with the WorldView-2 forward sight Depth extraction image that overall Kappa coefficient is bigger.

(2) the overall precision checking that the inverting of depth of water multiple-angle thinking is merged is analyzed

Source result before depth of water multi-angle inverting fusion results and fusion is made precision comparison, and each precision evaluation index obtained is in Table 2.

Table 2: the overall precision that depth of water multi-angle inverting is merged compares

Average relative error is followed successively by depth of water multi-angle inverting fusion evaluation, WorldView-2 forward sight Depth extraction image and WorldView-2 backsight Depth extraction image from small to large, value respectively 17.2%, 32.5% and 51.5%, comparing good forward sight image in stereogram, after fusion, the ensemble average relative error of image improves 15 percentage points. Mean absolute error also complies with the trend of average relative error, although the mean absolute error compared with forward sight Depth extraction image 0.9m, the mean absolute error of depth of water solid inverting fusion evaluation only has the reduction of 0.1m, but consider that the scope of the inverting depth of water is at 20m, error amount can reach sub-meter grade, and inversion accuracy is higher at last. And it being subject in image the effect of noise such as solar flare, BAIGUAN, the mean absolute error of the WorldView-2 backsight image Depth extraction obtained with the period is up to 2.2m.

As shown in Fig. 4 a, 4b, 4c, comparing this 3 scape depth of water image value at depth of water checkpoint place, no matter investigate from segmentation precision or inversion error, the WorldView-2 forward sight image used in experiment is better than backsight image in Depth extraction ability. But depth of water multi-angle inverting is merged by backsight image certain contribution, and the image precision after otherwise merging will not compare forward sight Depth extraction image the raising of such amplitude. As shown in Figure 5, it can be seen that image headwater depth level transition is better, and reef dish is clearly obvious.

(3) the segmentation precision test that the inverting of depth of water multiple-angle thinking is merged is analyzed

In order to better the effect of depth of water multi-angle inverting fusion experiment is evaluated, average relative error and the mean absolute error of analyzing 4 depth of water sections further are shown in Table 3.

Table 3: the segmentation error that depth of water multi-angle inverting is merged compares

In the depth of water section less than 2m, by depth of water multi-angle fusion evaluation, forward sight image, the order arrangement of backsight image, average relative error increases successively, and mean absolute error is that multiple increases. Compared with the image of source, the average relative error of depth of water multi-angle fusion evaluation reduces more than 100 percentage point, is 62.4%, and mean absolute error only has 0.3m. In the depth of water section more than 2m, two precision evaluation indexs of depth of water multi-angle fusion evaluation are all identical with WorldView-2 forward sight Depth extraction image, it is much better than the precision of backsight Depth extraction image, the above two are in the average relative error of 2-5m depth of water section and mean absolute error respectively 18.2% and 0.7m, and the latter is 86.4% and 3.1m. In the depth of water section of 5-10m and 10-20m, minimum average relative error is respectively less than 10%, and inversion accuracy is higher. Although in the deep water section more than 10m, minimum average B configuration absolute error is more than 1m, and for 1.4m, but except this depth of water section, the mean absolute error of depth of water multi-angle fusion evaluation and WorldView-2 forward sight Depth extraction image has all reached sub-meter grade.

Be can be seen that by segmentation precision analysis, on WorldView-2 forward sight Depth extraction image basis, depth of water multi-angle fusion evaluation improving mainly in the depth of water section less than 2m precision, namely, by the Decision fusion with backsight image, final result is learnt from other's strong points to offset one's weaknesses, obtains more optimizing the depth of water image that precision is higher.

The present invention is compared with existing inversion method, the present invention utilizes multiple-angle thinking to many-sided imaging capability under different sea situations, different substrate and different quality situation, abundant data message can be brought for RS Fathoming inverting, improve inversion accuracy, the marine sounding being particularly suited under complex situations shallow water area.

The technology contents of the not detailed description of the present invention is known technology.

Claims (3)

1. multiple-angle thinking depth of water Decision fusion inversion method, it is characterised in that comprise the following steps:

The first step: multi-spectrum remote sensing image is carried out pretreatment, obtains sea table reflectance;

Described pretreatment includes radiance conversion, atmospheric correction and solar flare and removes;

Second step: field measurement water depth value obtains and processes;

Obtain the bathymetric data of test block and corresponding latitude and longitude coordinates, the tidal height value measuring the moment is confirmed by tide table, bathymetric data correction is obtained the depth of water of theoretical depth datum level, further according to the acquisition moment of multi-spectrum remote sensing image, the bathymetric data of theoretical depth datum level is carried out the tidal correction of the instantaneous depth of water to obtain the instantaneous depth of water;

3rd step: single angle Depth extraction and depth of water segment identification;

According to the depth of water control point place depth of water and the relation between corresponding image picture element reflectance value, adopt multiband model to carry out statistical regression to go out, export the input that the parameter of this scape image Depth extraction merges as multi-angle Depth extraction, and multiband model is carried out parameter calibration, multiband model formation is as follows

$Z=A_0+\underset{i=1}{\overset{n}{\&Sigma;}}{A}_i{X}_i---\left(1\right)$

X_i=Ln (��_i-��_si)(2)

Wherein, Z is the depth of water, and n is the wave band number participating in inverting, A₀And A_iFor undetermined coefficient, ��_iIt is the i-th wave band reflectivity data, ��_siIt it is the reflectance at this wave band deep water place;

Depth of water control point being divided into multiple depth of water section as input, export average relative error and the credibility of each depth of water section, another fusion as multi-angle Depth extraction inputs, i.e. fusion parameters,

${\mathrm{\&delta;}}_k=\frac{1}{n}\underset{i=1}{\overset{n}{\&Sigma;}}\frac{|z_i-{z_i}^{\&prime;}|}{z_i}---\left(3\right)$

$L_k=0.5+{\left\{\underset{i=1}{\overset{n}{\&Sigma;}}{w}_i{\left(L_k^i-0.5\right)}^{\mathrm{\&alpha;}}\right\}}^{\frac{1}{\mathrm{\&alpha;}}}---\left(4\right)$

Wherein, k represents depth of water section, and n is depth of water checkpoint number, in formula 3, and ��_kIt is the average relative error of kth depth of water section, z_iIt is the measured value of i-th depth of water checkpoint, z_i' for its inverting value, in formula 4, L_kRepresent the credibility of kth depth of water section, w_iRepresent the relative importance of each information source, and meetIts formula is Representing the Kappa coefficient at this i scape image control point, n represents the number participating in inverting fusion evaluation, and �� is odd number,

Utilize fusion parameters and whole scape remote sensing image, calculate and obtain single angle Depth extraction result, and by its tidal correction to theoretical depth datum level, obtain depth of water segment identification image afterwards;

4th step: multi-angle Depth extraction merges;

The input single angle Depth extraction result, depth of water segment identification image and fusion parameters merged as multi-angle Depth extraction, carries out fusion by pixel, specifically includes:

(1) judging the depth of water segment identification that pixel is corresponding, next step computing of carrying out that pixel number in same depth of water section and this depth of water section is maximum, otherwise, according to (3rd) stepping row operation;

(2) judging the water depth value of pixel, if there being the pixel value of 2 or more than 2 identical, composing this value is final water depth value, if it is different, carry out next step computing;

(3) judging whether pixel is singular value in 5 �� 5 neighborhoods, selecting nonsingular value point assignment is final water depth value, if not singular point number has 2 or more than 2, carrying out next step computing, if being singular point, carrying out according to (5th) step;

(4) pixel average relative error in current depth of water section is compared, with pixel value corresponding to minimum error for final water depth value;

(5) intermediate value of pixel 5 �� 5 neighborhood is calculated, and with this neighborhood intermediate value for new pixel value, it is judged that its depth of water section belongs to, if identical, carries out according to (4th) step, otherwise carries out next step computing;

(6) the credibility size of pixel value place depth of water section is compared, with the big pixel value of credibility for final water depth value;

5th step: Depth extraction precision test;

Described precision test is the comparison of multi-angle inversion result after utilizing checkpoint to carry out the front single angle inversion result of fusion and merge, and after Depth extraction precision test completes, as the actual water depth value of remote sensing images, final water depth value is exported data.

2. multiple-angle thinking depth of water Decision fusion inversion method according to claim 1, it is characterised in that the spoke brightness transition in the described first step is that remote sensing image DN value is converted into spoke brightness value; Described solar flare is removed can adopt median method, averaging method or wavelet method; Described atmospheric correction can adopt FLAASH, dark pixel or 6S atmospheric correction method.

3. multiple-angle thinking depth of water Decision fusion inversion method according to claim 1, it is characterised in that the precision test in described 5th step includes average relative error that is overall and that divide different water depth section and mean absolute error.