CN109035223A - A kind of intelligent evaluation method for satellite remote sensing images availability - Google Patents
A kind of intelligent evaluation method for satellite remote sensing images availability Download PDFInfo
- Publication number
- CN109035223A CN109035223A CN201810755911.3A CN201810755911A CN109035223A CN 109035223 A CN109035223 A CN 109035223A CN 201810755911 A CN201810755911 A CN 201810755911A CN 109035223 A CN109035223 A CN 109035223A
- Authority
- CN
- China
- Prior art keywords
- availability
- remote sensing
- cloud
- image
- sensing image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 33
- 238000001514 detection method Methods 0.000 claims abstract description 12
- 238000013467 fragmentation Methods 0.000 claims abstract description 10
- 238000006062 fragmentation reaction Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 42
- 238000013210 evaluation model Methods 0.000 claims description 25
- 230000005284 excitation Effects 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 17
- 238000000034 method Methods 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 230000003044 adaptive effect Effects 0.000 claims description 10
- 238000012937 correction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 abstract 2
- 238000012545 processing Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 230000004075 alteration Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30192—Weather; Meteorology
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of intelligent evaluation method for satellite remote sensing images availability, includes the following steps: S1: image input；S2: inputting type of ground objects according to user demand and calculates the atural object affecting parameters of the type of ground objects；S3: cloud detection is carried out to the image of step S1 input；S4: cloud thickness, cloud degree of fragmentation, cloud coverage rate are calculated according to cloud detection result；S5: the cloud thickness calculated in step S4, cloud degree of fragmentation, cloud coverage rate input remote sensing images availability assessment models are calculated to the objective availability of corresponding remote sensing images；S6: the final availability rank of remote sensing images is calculated according to the atural object affecting parameters calculated in the objective availability and step S2 calculated in step S5；S7: corresponding remote sensing images are exported according to final availability rank.The invention proposes a kind of intelligent evaluation methods of remote sensing images availability to be designed in conjunction with the subjective impact for the objective factor and relevant user demand for influencing availability by reasonable step, for realizing that image availability is assessed.
Description
Technical Field
The invention relates to the field of remote sensing image application, in particular to an intelligent evaluation method for satellite remote sensing image availability.
Background
With the rapid development of remote sensing and space technologies, satellite remote sensing images are increasingly applied to various aspects such as earth resource investigation, natural disaster prediction and forecast, environmental pollution monitoring, satellite weather forecast, ground target identification and the like due to the advantages of large coverage area, strong time effectiveness, good data and geography comprehensiveness and the like. The acquisition and interpretation of high-quality satellite remote sensing images become one of the important supporting technologies for the development of national economy and national defense technologies.
Due to the cloud blocking problem, the subsequent image target identification, image classification and ground information extraction precision is reduced, even wrong results are caused, and the use of the remote sensing image is influenced. Meanwhile, the cloud coverage image also occupies a communication channel and a storage space, and wastes channel and ground equipment resources. Therefore, the satellite remote sensing image acquired by the sensor is difficult to be directly used for ground service due to the problem of interference on ground feature information generated in the information acquisition process such as cloud coverage, and the availability of the image needs to be evaluated and classified so as to fully utilize more effective ground feature information. Therefore, the availability evaluation of the remote sensing image is one of the key technologies for fully using the remote sensing image.
The method mainly aims at factors influencing the image availability, such as cloud thickness, distribution, area and shadow in the optical satellite remote sensing image, and the complexity of an underlying surface, and establishes an image availability evaluation model by using methods such as image processing, pattern recognition, probability statistics, artificial intelligence and the like, so that the availability evaluation result is ensured to be as same as the human interpretation result as possible, and the image data required by a user can be obtained by automatic classification instead of the manual interpretation method. In order to meet the requirement of intelligent evaluation, objective evaluation on the availability of an image is required to be carried out firstly, namely, before the image is provided for a user, the image is automatically classified and graded by using a machine intelligent method according to objective influence factors of the availability, so that the evaluation result is correct, and the visual interpretation work of professionals can be replaced.
Therefore, the objective evaluation method for the usability essentially aims to solve the problem of correctly classifying the images, and divides the image categories into a plurality of grades according to a specified standard, wherein each grade of the usability represents an image type. The problem belongs to the research category of pattern recognition, a computer is used for classifying certain image objects, and under the condition of minimum error probability, the recognition result is matched with an objective object as much as possible, namely, the computer is used for realizing the analysis, description, judgment and usability evaluation of people on the image. In order to achieve the purpose, the automatic classifier is trained by using the result of manual interpretation, and belongs to the category of supervised classification, namely, the classifier is trained and supervised by using a class image with known availability provided by a professional classifier in advance as a training sample, and then the image is classified.
At present, the China satellite image market does not form a perfect market system, the manual interpretation result is still the most important basis, the machine interpretation seems to hope to find the east and west with strong objective regularity from the samples, if the regularity of the samples is not strong, the memory function can be only exerted during prediction, and the effect of untouched samples is greatly reduced. The pertinence of the image use requirement is strong, and the influence of an emergent event is large. Although some basic factors influencing the image usability rating are reflected in the model by limited training samples, it is difficult to achieve comprehensive reflection.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to provide an intelligent evaluation method for the availability of a remote sensing image, which is used for realizing the evaluation of the availability of the remote sensing image through reasonable step design by combining objective factors influencing the availability and subjective influences related to user requirements.
Therefore, the invention provides the following technical scheme: an intelligent evaluation method for satellite remote sensing image availability is characterized by comprising the following steps:
s1: inputting an image;
s2, inputting the feature type according to the user requirement and calculating the feature influence parameter β of the feature type;
s3: performing cloud detection on the image input in step S1;
s4: calculating the cloud thickness, the cloud breaking degree and the cloud coverage rate according to the cloud detection result;
s5: inputting the cloud thickness, the cloud fragmentation degree and the cloud coverage rate calculated in the step S4 into a remote sensing image availability evaluation model to calculate objective availability output (cloud) of the corresponding remote sensing image;
s6: calculating a final availability level of the remote sensing image according to the objective availability calculated in the step S5 and the ground feature influence parameters calculated in the step S2;
wherein the final availability level of the remote sensing image is determined according to formula (1):
FA＝output(cloud)×β (1)
wherein, output (closed) is the objective availability of the remote sensing image, and β is a ground object influence parameter;
s7: and outputting the corresponding remote sensing image according to the final availability level of the remote sensing image calculated in the step S6.
The ground feature types comprise oceans, deserts, mountains, green lands and cities, and the input image is a satellite image of the middle-sized and small-sized bars.
Preferably, the feature influence parameter in step S2 is determined according to formula (2):
β＝1+flag×ρ-flag (2)
if the user needs to input the feature type, the flag is 1, and if the user does not need to input the feature type, the flag is 0;
when the flag is 1, detecting the ground feature type input by the user, and counting the total area S of the ground feature typeGroundStatistics of remoteTotal area S of all ground object types of sensed imageGeneral assemblyThen the coefficient of influence of the terrain isIf the user specifies that certain surface feature types are not desired,
preferably, the guest availability output (cloud) in step S5 is determined according to formula (3):
wherein, outputmf is the availability grade output by each fuzzy rule, WiI is an excitation intensity linear parameter, 1, 2, 3.. 27;
preferably, the image input in step S1 includes the steps of:
1) reading in an image;
2) correcting the color difference of the read image;
3) normalizing the resolution of the read-in image;
4) retaining the HSI component values after color difference correction, calculating the gray component gray through an RGB model, wherein the gray component gray is 0.299 xR +0.587 xG +0.114 xB, and inputting and storing the HSI three-component and the gray component gray as data for cloud detection;
r, G, B are the red, green, and blue components of the pixel, respectively;
the color difference correction is to convert the RGB model into HSI, retain H hue component, apply histogram equalization algorithm to S (saturation) and I (intensity) components, then store the new HSI component, and display the equalized color image,
resolution normalization is the reduction of the image by ceil rounding down functionsSmall processing, defining horizontal length hsize and vertical width vsize of input image, resolution is r m, then reducing horizontal length of image toHas a vertical width of
Preferably, the cloud thickness in step S4 is determined according to formula (4):
where ln is a natural logarithm function, Cb is the integrated luminance of the region, and Ac is the absolute contrast of the region (Ac ≠ 0).
Preferably, the cloud fragmentation degree in step S4 is determined according to formula (5):
CF＝PD/SV (5)
wherein PD is cloud area density, which refers to the number of connected areas of cloud speckles in unit area; SV is the variance of the area of the cloud area.
Preferably, the cloud coverage in step S4 is determined according to formula (6):
wherein S iscIs the area of the cloud coverage area, SzIs the total area of the image.
Preferably, the remote sensing image availability evaluation model is established by using an adaptive fuzzy neural network system by taking the availability level as a standard.
Preferably, the modeling of the remote sensing image availability evaluation model specifically includes the following steps:
step 1: determining 114 artificially labeled training samples, wherein each training sample has an output and three inputs, the three inputs are respectively CA, CF and CT, a fuzzy set is formed by CA, CF and CT, the fuzzy set is converted into membership functions through Gaussian function modeling, each input corresponds to 3 membership functions, and each input has 3 fuzzy intervals;
step 2: determining the number of membership functions as 9 according to the input of a training sample, wherein each membership function has 2 nonlinear parameters;
step 3: determining the number of fuzzy rules to be 27 according to the number of fuzzy intervals;
step 4: determining the number 27 of the excitation intensity linear parameters W according to the number of the fuzzy rules;
step 5: determining a network structure according to the number of membership functions, the number of fuzzy rules and the number of excitation intensity linear parameters;
step 6: training the network structure determined in the step5 by using an Anfis function in an MATLAB fuzzy toolbox to obtain 18 nonlinear parameters and 27 excitation intensity linear parameters of a membership function;
step 7: and substituting the 18 nonlinear parameters and 27 excitation intensity linear parameters which are obtained from step6 and used for obtaining the membership function into the network structure determined in step5 to obtain the availability evaluation model of the remote sensing image.
Preferably, the method further comprises a step S8, wherein the step S8 is a subsequent process, and includes adding an image learning library to the remote sensing image with the wrong scores output in the step S7, and determining a final usability level of the remote sensing image in a manual interpretation mode;
and feeding back the remote sensing image after artificial interpretation to the remote sensing image availability evaluation model, retraining the network structure, and correcting nonlinear parameters and linear parameters in the remote sensing image availability evaluation model.
Compared with the prior art, the invention has the following beneficial effects:
(1) the intelligent evaluation method for the availability of the satellite remote sensing image, provided by the invention, not only combines objective factors such as cloud thickness, cloud fragmentation degree and cloud coverage rate, but also considers influence factors of ground feature types, and is used for realizing the level evaluation of the availability of the remote sensing image through reasonable step design;
(2) the invention provides an intelligent evaluation method for satellite remote sensing image availability, which also comprises step S8 subsequent processing, including adding an image learning library to the output wrong remote sensing image, determining the final availability grade of the remote sensing image in a manual interpretation mode, feeding back the manually interpreted remote sensing image to a remote sensing image availability evaluation model in a characteristic parameter form, training a network structure again, continuously correcting nonlinear parameters and linear parameters in the remote sensing image availability evaluation model, so as to improve the functions of the remote sensing image in the long-term network use and updating process, continuously increase the accuracy of automatic interpretation and finally truly replace manual interpretation;
(3) the invention provides an objective evaluation model for the availability of a remote sensing image by using a self-adaptive fuzzy neural network, which is established by using three objective influence factors of the availability, namely cloud cover rate, cloud thickness and cloud fragmentation;
(4) the remote sensing image availability evaluation model is established by using an adaptive fuzzy neural network system by taking the availability grade as a standard, and requires that the evaluation result of a computer is consistent with the artificial result. Firstly, training samples are learned to obtain a series of parameters of a determined model, then, the usability of the image is tested and evaluated according to a model design classifier, and the artificial results are compared to give the closeness degree between the two. The higher the similarity rate, the lower the classification error rate, and the more effective the fuzzy neural network.
Drawings
FIG. 1 is a flowchart of a usability classification routine;
FIG. 2 is a flowchart of an image input procedure;
FIG. 3 is a diagram of five typical sample objects;
FIG. 4 is a diagram of a network architecture;
FIG. 5 is a diagram showing the result of image aberration correction;
FIG. 6 is a graph of the results of image resolution normalization;
FIG. 7 is an output gray scale graph;
FIG. 8 is a two-input one-output first order Sugeno fuzzy model diagram with two rules;
FIG. 9 is a diagram of an equivalent adaptive fuzzy neural network architecture.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings.
FIG. 1 is a flowchart of a usability classification ranking routine, as shown in FIG. 1;
an intelligent evaluation method for satellite remote sensing image availability is characterized by comprising the following steps:
s1: inputting an image;
s2, inputting the feature type according to the user requirement and calculating the feature influence parameter β of the feature type;
the feature influence parameter in the step S2 is determined according to the formula (2):
β＝1+flag×ρ-flag (2)
if the user needs to input the feature type, the flag is 1, and if the user does not need to input the feature type, the flag is 0;
when the flag is 1, detecting the ground feature type input by the user, and counting the total area S of the ground feature typeGroundCounting the total area S of all ground feature types of the remote sensing imageGeneral assemblyThen the coefficient of influence of the terrain isIf the user specifies that certain surface feature types are not desired,
s3: performing cloud detection according to the image input in step S1;
s4: calculating the cloud thickness, the cloud breaking degree and the cloud coverage rate according to the cloud detection result;
s5: inputting the cloud thickness, the cloud fragmentation degree and the cloud coverage rate calculated in the step S4 into a remote sensing image availability evaluation model to calculate objective availability output (cloud) of the corresponding remote sensing image;
in step S5, the guest availability output (cloud) is determined according to formula (3):
where outputmf (as shown in FIG. 4) is the availability level, W, output for each fuzzy ruleiI is an excitation intensity linear parameter, 1, 2, 3.. 27;
s6: calculating a final availability level of the remote sensing image according to the objective availability calculated in the step S5 and the ground feature influence parameters calculated in the step S2;
wherein the final availability level of the remote sensing image is determined according to formula (1):
FA＝output(cloud)×β (1)
wherein, output (closed) is the objective availability of the remote sensing image, and β is a ground object influence parameter.
The final availability level (FA) is determined by the objective availability of the remote sensing image and the type of the ground feature input by the user according to the actual demand.
S7: and outputting the corresponding remote sensing image according to the final availability level of the remote sensing image calculated in the step S6.
The method also comprises a step S8, wherein the step S8 is subsequent processing, and comprises the steps of adding an image learning library into the wrongly-divided remote sensing image output in the step S7, and determining the final usability level of the remote sensing image in a manual interpretation mode;
and feeding back the remote sensing image after artificial interpretation to the remote sensing image availability evaluation model, retraining the network structure, and correcting nonlinear parameters and linear parameters in the remote sensing image availability evaluation model.
As shown in fig. 2, fig. 2 is a flowchart of an image input procedure;
1) reading in an image;
2) correcting the color difference of the read image;
3) normalizing the resolution of the read-in image;
4) retaining the HSI component values after color difference correction, calculating the gray component gray through an RGB model, wherein the gray component gray is 0.299 xR +0.587 xG +0.114 xB, and inputting and storing the HSI three-component and the gray component gray as data for cloud detection; r, G, B are the red, green, and blue components of the pixel, respectively;
generally, after a series of preprocessing such as precise geometric correction and registration, a pseudo color image synthesized by images of three bands 2, 3 and 4 of five bands of a CCD camera is provided to a user, and the image storage format mainly includes BMP, JPEG and TIF. And respectively writing corresponding image reading programs according to the possible formats of the input images, and displaying the original images.
When the CCD is used for imaging, certain chromatic aberration occurs in an image due to the influence of light, the problem of color irregularity is presented, subsequent processing is influenced, and the chromatic aberration needs to be corrected. According to the HSI model used in the present document for processing data for multiple times, the RGB model is first converted into HSI, the H hue component is retained, the histogram equalization algorithm is applied to the S (saturation) and I (intensity) components, the new HSI component is then saved, and the equalized color image is displayed. The result of the correction is shown in figure 5,
in addition, with the development of remote sensing technology, the resolution of the image can be improved from 20m to 10m, 5m or even 1m, the processing speed is affected by the excessive image data amount, so the image is reduced, and if the horizontal length hsize and the vertical width vsize of the input image are set, and the resolution is r meters, the horizontal length of the reduced image is set asHas a vertical width ofceil is a floor function. The image subjected to the resolution normalization processing is shown in fig. 6.
The image processed as described above retains its HSI component values, and its gradation component gray is calculated by the RGB model to be 0.299 × R +0.587 × G +0.114 × B, and the HSI three-component and gradation component gray are stored as data inputs for cloud detection. The grayscale map output is shown in fig. 7.
The cloud thickness in step S4 is determined according to formula (4):
wherein ln is a natural logarithm function, Cb is the comprehensive brightness of the region, and Ac is the absolute contrast of the region (Ac ≠ 0);
cb is the cloud pixel integrated brightness calculated by the HSV model:
Cb(i)＝V(i)-S(i) (4.1)
wherein,
v is the brightness (Value) of the region, and S is the Saturation (Saturation) of the I region.
Ac represents that the satellite image contrast is the gray level difference degree representing the satellite image, the specific calculation method is as formula (3.8),
wherein p isiIs the gray level, s, corresponding to the peak value in the gray histogramiIs the gray level covered by the image.
The cloud fragmentation degree in step S4 is determined according to formula (5):
CF＝PD/SV (5)
wherein PD is cloud area density, which refers to the number of connected areas of cloud speckles in unit area; SV is the variance of the area of the cloud area.
The calculation methods of PD and SV are as follows:
wherein RN is the number of patches, SiIs the area of the ith cloud region, n is the number of the cloud regions, ScIs the area of the cloud coverage area, SGeneral assemblyIs the total area of the image.
The cloud coverage in step S4 is determined according to formula (6):
wherein S iscIs the area of the cloud coverage area, SzIs the total area of the image.
The remote sensing image availability evaluation model is established by using an adaptive fuzzy neural network system by taking an availability grade as a standard; and requires that the computer evaluation result be consistent with the manual result. Firstly, training samples are learned to obtain a series of parameters of a determined model, then, the usability of the image is tested and evaluated according to a model design classifier, and the artificial results are compared to give the closeness degree between the two. The higher the similarity rate, the lower the classification error rate, and the more effective the fuzzy neural network.
Dividing the availability of the image into six levels of 0, 1, 2, 3, 4 and 5; according to the experience classification method of workers of the national resource satellite application center, aiming at the user requirements in the remote sensing image productization process, images are observed for a long time, the availability of the images is subjectively evaluated, and under the condition that the users are unified and the images do not need to contain surface feature types, the scores are counted and averaged to obtain an evaluation result.
Adaptive neural-Fuzzy Inference System (ANFIS) was Jang in the early 90 s[88]The proposed architecture, which is a fuzzy system implemented with adaptive networks. The system is based on a data modeling method, and the fuzzy membership function and the fuzzy rule are obtained through learning of known data and are not given based on experience or intuition. The method is very suitable for systems with complex characteristics and incomplete solution.
ANFIS is a fuzzy neural network based on T-S (Takagi-Sugeno) inference, which can be expressed in the form of: if x1isA and x2isB then y＝f(x1,x2). For simplicity, assume that the fuzzy inference system under consideration has two outputsInto x1And x2And outputting y singly. For the first order Sugeno fuzzy model, the common rule set with two fuzzy "if-then" rules is as follows:
rule 1: if x1Is A1and x2Is B1Then f1＝p1x1+q1x2+r1，
Rule 2: if x1Is A2and x2Is B2Then f2＝p2x1+q2x2+r2。
The reasoning mechanism of this Sugeno model is explained as FIG. 8; FIG. 8 is a two-input one-output first order Sugeno fuzzy model diagram with two rules;
as shown in fig. 9, fig. 9 is a diagram of an equivalent adaptive fuzzy neural network structure, in which the first layer: the output of each node function is an input pair fuzzy set (A)1,A2,B1,B2) Degree of membership of:
the membership may be given by any parameterized membership function, such as a bell function, Sigmoid function, gaussian function, etc.
A second layer: each node of this layer is a fixed node labeled Π, whose output is the product of all the input signals:
the output of each node represents the excitation strength of a rule. In general, any other T-norm operator that performs a fuzzy AND may be used for the node functions of this layer.
And a third layer: the excitation strength W of the corresponding rule is calculated by each nodei(excitation intensity of ith rule) to the sum of all rule W (excitation intensity of all rules) values:
O3,i＝Wi/W (3.14)
for convenience, the output of this layer is referred to as the normalized excitation intensity.
A fourth layer: each node of the hierarchy is an adaptive node having a parameterized node function
O4,i＝O3,ifi＝O3,i(pix1+qix2+ri) i＝1，2 (3.15)
In the formula O3,iIs the normalized excitation intensity, p, coming from layer 3i,qi,riIs the parameter set for that node. The parameters of this layer are called conclusion parameters.
The fifth level is a fixed node which calculates the sum of all transmitted signals as the total output f (x)1,x2)。
It can be seen that the adaptive fuzzy neural network is a five-layer forward network functionally equivalent to the Sugeno fuzzy model. Each layer of nodes of the network belong to fuzzy neurons with different functions.
The modeling of the remote sensing image availability evaluation model specifically comprises the following steps:
step 1: determining 114 artificially labeled training samples, wherein each training sample has an output and three inputs, the three inputs are respectively CA, CF and CT, a fuzzy set is formed by CA, CF and CT, the fuzzy set is converted into membership functions through a Gaussian function, each input corresponds to 3 membership functions, and each input has 3 fuzzy intervals;
step 2: determining the number of membership functions as 9 according to the input of a training sample, wherein each membership function has 2 nonlinear parameters;
step 3: determining the number of fuzzy rules to be 27 according to the number of fuzzy intervals;
step 4: determining the number 27 of the excitation intensity linear parameters W according to the number of the fuzzy rules;
step 5: determining a network structure according to the number of membership functions, the number of fuzzy rules and the number of excitation intensity linear parameters;
step 6: training the network structure determined in the step5 by using an Anfis function in an MATLAB fuzzy toolbox to obtain 18 nonlinear parameters and 27 excitation intensity linear parameters of a membership function;
step 7: and substituting the 18 nonlinear parameters and 27 excitation intensity linear parameters which are obtained from step6 and used for obtaining the membership function into the network structure determined in step5 to obtain the availability evaluation model of the remote sensing image.
The selection of the number of fuzzy intervals in Step3 is determined according to experience of a designer and multiple tests, and specifically comprises the following steps: according to the network structure, the number of fuzzy intervals is selected, the influence of the number of the fuzzy intervals on the network performance is large, and the interval division is too fine, so that the network learning time is too long, and even convergence cannot be realized; too coarse a separation of intervals may result in poor fault tolerance of the network.
The selection of the number of fuzzy intervals is a very complicated problem, and is often determined according to experience of designers and a plurality of tests, and an ideal analytic expression for calculating the number is not available at present. But reference may be made to the following formula (7):
wherein L is the number of fuzzy intervals, m is the number of input nodes, n is the number of output nodes, and c is a constant between 1 and 10.
According to the formula, L can be selected to be a value between 3 and 12, repeated trials show that when L is 3, the network performance can be optimal, when L is continuously increased, the network performance is not obviously improved, but is reduced, which shows that the number of fuzzy intervals is better if the number of fuzzy intervals is larger, so that the number of fuzzy intervals of the finally determined three input features is [333],
an intelligent evaluation method for satellite remote sensing image availability further comprises a step S7, wherein the step S7 is follow-up processing, and comprises the steps of adding an image learning library to an output wrongly-divided remote sensing image, and determining the final availability level of the remote sensing image in a manual interpretation mode;
and feeding back the artificially interpreted remote sensing image to the remote sensing image availability evaluation model in a characteristic parameter form, retraining the network structure, and correcting nonlinear parameters and linear parameters in the remote sensing image availability evaluation model.
The above-described series of detailed descriptions are merely specific to possible embodiments of the present invention, and they are not intended to limit the scope of the present invention, and various changes made without departing from the gist of the present invention within the knowledge of those skilled in the art are within the scope of the present invention.
Claims (10)
1. An intelligent evaluation method for satellite remote sensing image availability is characterized by comprising the following steps:
s1: inputting an image;
s2, inputting the feature type according to the user requirement and calculating the feature influence parameter β of the feature type;
s3: performing cloud detection on the image input in step S1;
s4: calculating the cloud thickness, the cloud breaking degree and the cloud coverage rate according to the cloud detection result;
s5: inputting the cloud thickness, the cloud fragmentation degree and the cloud coverage rate calculated in the step S4 into a remote sensing image availability evaluation model to calculate objective availability output (cloud) of the corresponding remote sensing image;
s6: calculating a final availability level of the remote sensing image according to the objective availability calculated in the step S5 and the ground feature influence parameters calculated in the step S2;
wherein the final availability level of the remote sensing image is determined according to formula (1):
FA＝output(cloud)×β (1)
wherein, output (closed) is the objective availability of the remote sensing image, and β is a ground object influence parameter;
s7: and outputting the corresponding remote sensing image according to the final availability level of the remote sensing image calculated in the step S6.
2. The intelligent evaluation method according to claim 1, wherein the feature influence parameter in step S2 is determined according to formula (2):
β＝1+flag×ρ-flag (2)
if the user needs to input the feature type, the flag is 1, and if the user does not need to input the feature type, the flag is 0;
when the flag is 1, detecting the ground feature type input by the user, and counting the total area S of the ground feature typeGroundCounting the total area S of all ground feature types of the remote sensing imageGeneral assemblyThen the coefficient of influence of the terrain isIf the user specifies that certain surface feature types are not desired,
3. the intelligent evaluation method according to claim 1, wherein the guest availability output (cloud) in step S5 is determined according to formula (3):
wherein, outputmf is the availability grade output by each fuzzy rule, WiI is an excitation intensity linear parameter, 1, 2, 3.
4. The intelligent evaluation method according to claim 1, wherein the image input in the step S1 includes the steps of:
1) reading in an image;
2) correcting the color difference of the read image;
3) normalizing the resolution of the read-in image;
4) retaining the HSI component values after color difference correction, calculating the gray component gray through an RGB model, wherein the gray component gray is 0.299 xR +0.587 xG +0.114 xB, and inputting and storing the HSI three-component and the gray component gray as data for cloud detection;
r, G, B are the red, green, and blue components of the pixel, respectively;
the color difference correction is to convert the RGB model into HSI, reserve H hue component, apply histogram equalization algorithm to S (saturation) and I (intensity) component, then store new HSI component, and display equalized color image;
the resolution normalization is to reduce the image by ceil rounding function, define the horizontal length hsize and vertical width vsize of the input image, and reduce the horizontal length of the image to r mHas a vertical width of
5. The intelligent evaluation method according to claim 1, wherein the cloud thickness in step S4 is determined according to formula (4):
where ln is a natural logarithm function, Cb is the integrated luminance of the region, and Ac is the absolute contrast of the region (Ac ≠ 0).
6. The intelligent evaluation method according to claim 1, wherein the cloud fragmentation degree in step S4 is determined according to formula (5):
CF＝PD/SV (5)
wherein PD is cloud area density, which refers to the number of connected areas of cloud speckles in unit area; SV is the variance of the area of the cloud area.
7. The intelligent evaluation method according to claim 1, wherein the cloud coverage in the step S4 is determined according to formula (6):
wherein S iscIs the area of the cloud coverage area, SzIs the total area of the image.
8. The intelligent evaluation method according to claim 1, wherein the remote sensing image availability evaluation model is established by using an adaptive fuzzy neural network system with the availability level as a standard.
9. The intelligent evaluation method according to claim 8, wherein the modeling of the remote sensing image availability evaluation model specifically comprises the steps of:
step 1: determining 114 artificially labeled training samples, wherein each training sample has an output and three inputs, the three inputs are respectively CA, CF and CT, a fuzzy set is formed by CA, CF and CT, the fuzzy set is converted into membership functions through a Gaussian function, each input corresponds to 3 membership functions, and each input has 3 fuzzy intervals;
step 2: determining the number of membership functions as 9 according to the input of a training sample, wherein each membership function has 2 nonlinear parameters;
step 3: determining the number of fuzzy rules to be 27 according to the number of fuzzy intervals;
step 4: determining the number 27 of the excitation intensity linear parameters W according to the number of the fuzzy rules;
step 5: determining a network structure according to the number of membership functions, the number of fuzzy rules and the number of excitation intensity linear parameters;
step 6: training the network structure determined in the step5 by using an Anfis function in an MATLAB fuzzy toolbox to obtain 18 nonlinear parameters and 27 excitation intensity linear parameters of a membership function;
step 7: and substituting the 18 nonlinear parameters and 27 excitation intensity linear parameters which are obtained from step6 and used for obtaining the membership function into the network structure determined in step5 to obtain the availability evaluation model of the remote sensing image.
10. The intelligent evaluation method according to claim 1, further comprising a step S8, wherein the step S8 is a subsequent process, which comprises adding the wrongly-classified remote sensing image outputted in the step S7 to an image learning library, and determining a final usability level of the remote sensing image in a manual interpretation manner;
and feeding back the remote sensing image after artificial interpretation to the remote sensing image availability evaluation model, retraining the network structure, and correcting nonlinear parameters and linear parameters in the remote sensing image availability evaluation model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810755911.3A CN109035223A (en) | 2018-07-11 | 2018-07-11 | A kind of intelligent evaluation method for satellite remote sensing images availability |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810755911.3A CN109035223A (en) | 2018-07-11 | 2018-07-11 | A kind of intelligent evaluation method for satellite remote sensing images availability |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109035223A true CN109035223A (en) | 2018-12-18 |
Family
ID=64640958
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810755911.3A Pending CN109035223A (en) | 2018-07-11 | 2018-07-11 | A kind of intelligent evaluation method for satellite remote sensing images availability |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109035223A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826526A (en) * | 2019-11-19 | 2020-02-21 | 上海无线电设备研究所 | Method for cloud detection radar to identify clouds |
| CN110889840A (en) * | 2019-11-28 | 2020-03-17 | 航天恒星科技有限公司 | Effectiveness detection method of high-resolution 6 # remote sensing satellite data for ground object target |
| CN115082452A (en) * | 2022-07-26 | 2022-09-20 | 北京数慧时空信息技术有限公司 | Cloud and shadow based quantitative evaluation method for quality of remote sensing image |
| CN116612048A (en) * | 2023-07-17 | 2023-08-18 | 山东产研卫星信息技术产业研究院有限公司 | Method and system for deblurring optical satellite remote sensing image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103604421A (en) * | 2013-11-25 | 2014-02-26 | 上海航天测控通信研究所 | Cloud detection method based on satellite-borne hyperspectral image with spectral reflectance characteristic |
| CN105260729A (en) * | 2015-11-20 | 2016-01-20 | 武汉大学 | Satellite remote sensing image cloud amount calculation method on the basis of random forest |
-
2018
- 2018-07-11 CN CN201810755911.3A patent/CN109035223A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103604421A (en) * | 2013-11-25 | 2014-02-26 | 上海航天测控通信研究所 | Cloud detection method based on satellite-borne hyperspectral image with spectral reflectance characteristic |
| CN105260729A (en) * | 2015-11-20 | 2016-01-20 | 武汉大学 | Satellite remote sensing image cloud amount calculation method on the basis of random forest |
Non-Patent Citations (3)
| Title |
|---|
| 吴兴华等: "基于密度自适应粒子群优化的图像可用度分类", 《计算机仿真》 * |
| 文档收集: "《卫星遥感图像可用度智能评估方法研究》", 《道客巴巴 网址：HTTPS://WWW.DOC88.COM/P-9146604262080.HTML》 * |
| 郑红等: "基于云特征的遥感图像可用度影响因素分析", 《北京航天航空大学学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826526A (en) * | 2019-11-19 | 2020-02-21 | 上海无线电设备研究所 | Method for cloud detection radar to identify clouds |
| CN110889840A (en) * | 2019-11-28 | 2020-03-17 | 航天恒星科技有限公司 | Effectiveness detection method of high-resolution 6 # remote sensing satellite data for ground object target |
| CN115082452A (en) * | 2022-07-26 | 2022-09-20 | 北京数慧时空信息技术有限公司 | Cloud and shadow based quantitative evaluation method for quality of remote sensing image |
| CN115082452B (en) * | 2022-07-26 | 2022-11-04 | 北京数慧时空信息技术有限公司 | Cloud and shadow based quantitative evaluation method for quality of remote sensing image |
| CN116612048A (en) * | 2023-07-17 | 2023-08-18 | 山东产研卫星信息技术产业研究院有限公司 | Method and system for deblurring optical satellite remote sensing image |
| CN116612048B (en) * | 2023-07-17 | 2023-09-26 | 山东产研卫星信息技术产业研究院有限公司 | Method and system for deblurring optical satellite remote sensing image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11521380B2 (en) | Shadow and cloud masking for remote sensing images in agriculture applications using a multilayer perceptron | |
| CN104598908B (en) | A kind of crops leaf diseases recognition methods | |
| CN108615071B (en) | Model testing method and device | |
| CN110458077B (en) | Vehicle color identification method and system | |
| CN109035223A (en) | A kind of intelligent evaluation method for satellite remote sensing images availability | |
| CN111553240B (en) | Corn disease condition grading method and system and computer equipment | |
| CN109063754B (en) | Remote sensing image multi-feature joint classification method based on OpenStreetMap | |
| US20200250427A1 (en) | Shadow and cloud masking for agriculture applications using convolutional neural networks | |
| CN109544497A (en) | Image interfusion method and electronic equipment for transmission line faultlocating | |
| CN113344475B (en) | Transformer bushing defect identification method and system based on sequence modal decomposition | |
| CN111753805A (en) | Method and device for detecting wearing of safety helmet | |
| CN112417981B (en) | Efficient recognition method for complex battlefield environment targets based on improved FasterR-CNN | |
| CN111507426A (en) | No-reference image quality grading evaluation method and device based on visual fusion characteristics | |
| CN109886146B (en) | Flood information remote sensing intelligent acquisition method and device based on machine vision detection | |
| Pan et al. | Mapping cropland distributions using a hard and soft classification model | |
| CN110084284A (en) | Target detection and secondary classification algorithm and device based on region convolutional neural networks | |
| CN111079807A (en) | Ground object classification method and device | |
| Sowmya et al. | Land cover classification using reformed fuzzy C-means | |
| CN113361530A (en) | Image semantic accurate segmentation and optimization method using interaction means | |
| Su | [Retracted] Data Research on Tobacco Leaf Image Collection Based on Computer Vision Sensor | |
| CN116343048A (en) | Accurate land block boundary extraction method and system for plain crop type complex region | |
| Cui | Research on garden landscape reconstruction based on geographic information system under the background of deep learning | |
| Granados-López et al. | Pixel‐Based Image Processing for CIE Standard Sky Classification through ANN | |
| CN117853926B (en) | Building detection method and system based on artificial neural network classification | |
| Jasim et al. | Object-based Classification of Natural Scenes Using Machine Learning Methods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181218 |