CN107688782A - Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image - Google Patents
Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image Download PDFInfo
- Publication number
- CN107688782A CN107688782A CN201710727161.4A CN201710727161A CN107688782A CN 107688782 A CN107688782 A CN 107688782A CN 201710727161 A CN201710727161 A CN 201710727161A CN 107688782 A CN107688782 A CN 107688782A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- tank
- circle
- oil tank
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/68—Analysis of geometric attributes of symmetry
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30172—Centreline of tubular or elongated structure
Abstract
The present invention relates to a kind of oil tank detection based on high-resolution optical remote sensing image and reserve analysis method, belong to remote sensing image processing and analysis technical field, comprise the following steps:First, using improved ELSD (Ellipse and Line Segment Detector) and K means clustering algorithms, candidate's oil tank target is extracted;Then, DCNN (Deep Convolutional Neural Network) model candidate target input trained, accurate testing result is obtained, and realizes the classification to flat-top tank and arch-roof tank;Finally, using oil tank testing result and the symmetrical feature of flat-top tank surrounding shadow, calculate the stocking rate of flat-top tank and reserve analysis is carried out to it;In addition, being based on vision offset caused by shooting angle, reservoir data is calculated manually, is mutually authenticated with experimental data.The present invention quickly can accurately detect oil reserve reservoir area oil tank and carry out reserve analysis to flat-top tank, have great importance to grasping national economic strength, strategic decision, performing the information such as trend.
Description
Technical field
The invention belongs to remote sensing image processing and analysis technical field, and in particular to one kind is based on high-resolution optical
The oil tank detection of remote sensing images and reserve analysis method.
Background technology
With the development and progress of remote sensing satellite correlation technique, resolution ratio, the filming frequency all more and more highers of satellite image,
More, more reliable data is provided for image interpretation work to support.In remote sensing ground object target detection field, more mesh is studied
Indicate aircraft, steamer, airport, oil depot, building area etc..The oil storage tank facility indispensable as strategic oil reserve, in safety
The fields such as monitoring, disaster prevention, intelligence analysis possess very important status.The extraction of oil reservoir area tank information and analysis pair
The information such as national economic strength, strategic decision, execution trend are grasped to have great importance.
Traditional oil tank detection method is based on manual features, while also gradually from conventional method to deep learning method mistake
Cross.There is method[1]Propose and be applied to use by improved Canny rim detections, Hough transform and Fast template matching algorithm
The image of Brovey conversion fusion method processing, reach 85% accuracy rate;There is method[2]It is proposed is based on synthetic aperture radar figure
The oil tank detection method of director circle shade and highlighted arc as in;There is method[3,4]The detection algorithm based on symmetrical feature is proposed, can be with
Effectively alleviate rapidly radially symmetry transformation (FRST)[5]Some drawbacks;There is method[6]Propose and be based on improved Hough transform
Method, orientation and Weighted H ough Voting Algorithms;There is method[7]Propose the handling process of more systematization:First by EMHC
Conspicuousness model inspection goes out potential target region, forms conspicuousness collection of illustrative plates;Then examined on conspicuousness collection of illustrative plates using Hough transform
Measure circular target;Finally verify whether these circular targets belong to storage tank using SVM algorithm;Also there is method[8]The processing of proposition
Flow is more ripe, is broadly divided into three steps:Candidate Set selection, feature extraction and classification checking.The method use it is more quick,
Efficient ELSD (Ellipse and Line Segment Detector) algorithm, and certain improvement is made to it.
Following problem be present in existing method:Hough transform and its related algorithm are mostly based on, is stranded with parameter setting
It is difficult, pair radius is sensitive, is only applicable to that contrast is clear and the shortcomings of profile complete image;The factors such as shade, texture are not considered
Interference to positioning accuracy;It is concerned only with test problems in itself, does not carry out more deep analysis.
It is described pertinent literature below:
[1]W.Zhang,H.Zhang,C.Wang and T.Wu.Automatic oil tank detection
algorithm based on remote sensing image fusion[C].Proceedings of IEEE IGARSS,
2005,6(1):3956–3958.
[2]Xu H,Chen W,Sun B,et al.Oil tank detection in synthetic aperture
radar images based on quasi-circular shadow and highlighting arcs[J].Journal
of Applied Remote Sensing,2014,8(1):397–398.
[3]Ok A O,E.Circular oil tank detection from panchromatic
satellite images:a new automated approach[J].IEEE Geoscience and Remote
Sensing Letters,2015,12(6):1347-1351.
[4]Ok A O,Baseski E.Automated detection of oil depots from high
resolution images:A new perspective[J].ISPRS Annals of the Photogrammetry,
Remote Sensing and Spatial Information Sciences,2015,2(3):149-156.
[5]Loy G,Zelinsky A.Fast radial symmetry for detecting points of
interest[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2003,
25(8):959-973.
[6]Zhao W,Yang H,Shen Z,et al.Oil tanks extraction from high
resolution imagery using a directional and weighted hough voting method[J]
.Journal of the Indian Society of Remote Sensing,2015,43(3):539-549.
[7]Cai X,Sui H,Ruipeng L V,et al.Automatic circular oil tank
detection in high-resolution optical image based on visual saliency and Hough
transform[C].Electronics,Computer and Applications,2014 IEEE Workshop
on.IEEE,2014:408-411.
[8]Zhang L,Shi Z,Wu J.A hierarchical oil tank detector with deep
surrounding features for high-resolution optical satellite imagery[J].IEEE
Journal of Selected Topics in Applied Earth Observations and Remote Sensing,
2015,8(10):4895-4909.
The content of the invention
The technology of the present invention solves problem:In view of the deficienciess of the prior art, propose a kind of distant based on high-resolution optical
Feel oil tank detection and the reserve analysis method of image, realize the automatic detection and Accurate classification to oil reserve reservoir area oil tank;Base
In oil tank testing result and the symmetrical feature of flat-top tank surrounding shadow, calculate the stocking rate of flat-top tank and such oil tank is stored up
Amount analysis;Based on vision offset caused by shooting angle, reservoir data is calculated manually, is mutually authenticated with experimental data.
The technical solution of the present invention:A kind of oil tank detection and reserve analysis based on high-resolution optical remote sensing image
Method, it is characterised in that comprise the following steps:
Step 1:Gather and mark the remote sensing image data of the target containing oil tank, split data into training set, checking collection
And test set;
Step 2:Based on improved ELSD (Ellipse and Line Segment Detector) algorithm, step 1 is obtained
To training set, checking collection and test set extract candidate target respectively;
Step 3:Training set, the checking collection candidate target obtained using step 2 is trained and optimizes DCNN (Deep
Convolutional Neural Network) model, the DCNN that the test set candidate target input that step 2 obtains has been trained
Model, exports target classification result, and the target classification result includes flat-top tank, arch-roof tank and the class of background three;
Step 4:Using the symmetrical feature of flat-top tank surrounding shadow, based on point to the shade symmetry analysis method compared, carry
Make even and push up the optimal axis of symmetry in tank surrounding shadow region;
Step 5:The axis of symmetry obtained according to shade thickness and step 4, the stocking rate of flat-top tank is calculated;
Step 6:True reserve data are difficult to obtain, and based on vision offset caused by shooting angle, calculate manually
Reservoir data, it is mutually authenticated with real data.
Preferably, the specific implementation of step 2 includes following sub-step:
Step 2.1:Whether two discrete circular arcs of checking belong to same annulus, that is, determine the membership of circular arc,
Provided by following three formula:
Cenr, Cenc represent the row, column coordinate in the center of circle of some circular arc respectively, and r represents the radius of circular arc, subscript a, b table
Show two different circular arcs, θ, σ are small constant, represent specific threshold value, and formula (2) is used to confirm whether two circular arcs have
The same center of circle;Formula (3) is used to confirm whether two circular arcs have identical radius;
Step 2.2:The number of pixels for all circular arc fragments corresponding to same annulus that calculating detects accounts for the annulus should
There is the ratio of number of pixels;
lcircleRepresent all pixels points on a certain annulus, kx(circle) circle detected on corresponding annulus is represented
Total pixel number of arc, RcircleValue between 0 to 1, and its value, closer to 1, respective regions are more likely to be mesh
Mark;Set certain threshold value R0, the candidate region after being filtered.
Step 2.3:It is grouped according to the candidate region that Jaccard similarities obtain step 2.2, is gathered equivalent to space
Class, each packet correspond to unique target;
Area represents the area of some circle detected, and subscript c, d represents the different candidate regions that step 2.2 obtains
Domain, when calculating in order to simple, directly with circular external square cartographic represenation of area;
Step 2.4:Two class clusters are carried out for each packet, then according to space length, are determined most for each target
Unique candidate region eventually, so far extraction obtain candidate target.
Each cluster the mean circle-center coordinate of all candidate regions and radius corresponding to it represent, i.e., form for (Cenr,
Cenc, r) triple.If the triple of two clusters meets formula (2), formula (3) is but unsatisfactory for, illustrates two clusters
The center of circle is consistent, and radius is inconsistent, then the triple of the larger cluster of radius is using as the spatial information of respective objects;It is because this
In the case of, the cluster with shorter radius is likely to as caused by the inside circular profile on flat-top tank top, and radius is smaller
Cluster should be rejected;In addition, i.e. the center of circle is inconsistent, then quantity less cluster in candidate region is likely to be produced by shade
, now the triple of the larger cluster of radix is using as the spatial information of respective objects.
Preferably, the specific implementation of step 4 includes following sub-step:
Step 4.1:The center of circle of tank body cross section is set to fixing point, institute on circular arc is traveled through a little, on each circular arc
Point and the center of circle all uniquely determine straight line;
Step 4.2:Using the symmetrical feature of flat-top tank surrounding shadow, for each point in circle, verify the point on step
Whether the pixel value of the symmetric points of rapid 4.1 gained straight line is consistent with the pixel value;
Step 4.3:For every straight line, the number of the interior inconsistent logarithm on the straight line of statistics circle;
Step 4.4:According to step 4.3, the minimum straight line of inconsistent logarithm is obtained, it is defeated as the final axis of symmetry
Go out.
Preferably, the specific implementation of step 5 includes following sub-step:
Step 5.1:As shown in figure 5, finding the intersecting point coordinate of the axis of symmetry obtained in step 4 and circular arc, and calculate two
The shadows pixels on the axis of symmetry between individual intersection point are counted out, as partial phantom thickness lin;
Step 5.2:The incident direction of light is determined according to the offset direction of shade in step 5.1, and according to incident direction
The position of shade outside tank body is determined, calculates the shade thickness l that backlight side dash area intersects with symmetry axisout;
Step 5.3:According to parallel projection principle and triangle correspondence theorem, step 5.1 and step 5.2 two parts shade are thick
The ratio of degreeEqual to the vacancy rate of oil tankSo as to obtain the stocking rate of oil tankWherein h1And h2Represent respectively
Tank deck is relative to ground and the difference in height of flat-top.
Preferably, the specific implementation of step 6 includes following sub-step:
Step 6.1:As shown in fig. 6, due to the presence of shooting angle, edge, the edge and tank body of floating roof at the top of tank body
The edge of bottom staggers on two dimensional surface, and artificial measurement obtains corresponding shade thickness d in Fig. 71、d2;
Step 6.2:According to triangle correspondence theorem, the shade thickness d in step 6.11、d2RatioEqual to oil tank
Vacancy rateObtain another stocking rate dataAs experimental result true value, for carry out experimental evaluation with than
Compared with;
Step 6.3:When the true value that experimental result and step 6.2 are calculated compares, using root-mean-square error as commenting
Price card is accurate, and error is smaller, represents that experimental result is more reliable.
rreserveRepresent experimental result, rgroundArtificial result of calculation is represented, m represents sample size, and err represents root mean square
Error, represent the similarity degree between data.
The present invention compared with prior art the advantages of and good effect it is as follows:
(1) original false-alarm filter method is improved, the method for being depended on rate based on pixel number originally is changed to
Based on the method for circular arc radian accounting, solves the problems, such as original method to circular arc thickness-sensitive;The method of utilization space cluster,
Circular arc caused by filtering shade and texture, improve the accuracy of target positioning;Based on from labeled data train and optimize DCNN
Model, rather than directly used as feature extractor, enhance the applicability of model;Refined point on the basis of oil tank detection
Class, the classification to flat-top tank and arch-roof tank is realized, be easy to follow-up further reserve analysis.Oil tank detection of the present invention
Algorithm rapid extraction oil tank target, its detection and classification performance can be better than existing method automatically.
(2) it is theoretical based on oil tank testing result and shade symmetry analysis, the stocking rate of flat-top tank is calculated, and to such
Oil tank carries out the reserve analysis of heuristic.Reserve analysis method of the present invention has carried out deep spy to the research of oil tank class
Rope, gained stocking rate have higher accuracy.
(3) vision offset caused by shooting angle is based on, reservoir data is calculated manually, is mutually tested with experimental data
Card, solves the problems, such as the model evaluation in the case of True Data is difficult to obtain.
Brief description of the drawings
Fig. 1 is the implementation process figure of the present invention;
Fig. 2 is that the candidate target based on ELSD extracts flow chart;
Fig. 3 is DCNN network structures;
Fig. 4 is flat-top tank shade axis of symmetry analysis chart, wherein (a) is original image, (b) is the shade of extraction, and (c) is
Shade symmetry axis;
Fig. 5 is that flat-top tank shade forms exemplary plot;
Fig. 6 is vision offset schematic diagram caused by shooting angle;
Fig. 7 is that visual angle caused by shooting angle offsets schematic diagram.
Embodiment
With reference to embodiment and Figure of description, the embodiment of the present invention is described in detail.Retouch in this place
The embodiment stated is merely to illustrate and explain the present invention, and is not intended to limit the present invention.
As shown in figure 1, oil tank detection and reserve analysis side proposed by the present invention based on high-resolution optical remote sensing image
Method, comprise the following steps:First, using improved ELSD (Ellipse and Line Segment Detector) and K-
Means clustering algorithms, extract candidate's oil tank target;Then, DCNN (Deep candidate target input trained
Convolutional Neural Network) model, accurate testing result is obtained, and realize and flat-top tank and arch-roof tank are divided
Class;Finally, using oil tank testing result and the symmetrical feature of flat-top tank surrounding shadow, calculate the stocking rate of flat-top tank and it is entered
Row reserve analysis;In addition, being based on vision offset caused by shooting angle, reservoir data is calculated manually, with experimental data
It is mutually authenticated.The present invention quickly can accurately detect oil reserve reservoir area oil tank and carry out reserve analysis to flat-top tank, to grasping
The economic strength of country, strategic decision, perform the information such as trend and have great importance.
The following detailed description of.
Step 1:Gather and mark the remote sensing image data of the target containing oil tank, split data into training set, checking collection
And test set;
When it is implemented, training data should cover situation as much as possible, to ensure the completeness of training set.
Step 2:Based on improved ELSD algorithms, candidate's oil tank target is extracted;
Described ELSD algorithms refer to the joint detector of a line segment without ginseng and elliptic arc.It follows three-wave-length side
Case:Feature Candidate Set is determined by heuristic search;Removal can not be verified to obtained Candidate Set based on control methods
It can be the candidate item of line segment or circular arc;Based on given multiple feature races, optimal geometric interpretation is selected by model.Change
Method after entering is that step card is further added by the basis of former algorithm, fully improves and calculates performance.
Step 3:Train and optimize DCNN models, candidate target to be measured is inputted into training network, exports target classification knot
Fruit, including flat-top tank, arch-roof tank and the class of background three;
Described DCNN models include three convolutional layers (Convolutional layer), three pond layer (Pooling
Layer), a full articulamentum (Inner product layer) and softmax (Softmax layer) layer, they it
Between annexation it is as shown in Figure 3.Pond layer includes a maximum pond layer (Max pooling layer) and two average ponds
Change layer (Average pooling layer).Maximum pondization can be effectively reduced computation complexity;And average pondization can protect
More local messages are stayed, information is reduced and loses.
Step 4:Using the symmetrical feature of flat-top tank surrounding shadow, based on point to the shade symmetry analysis method compared, carry
Take the optimal axis of symmetry of shadow region;
Described shade symmetry analysis method depends on two reasonings:(1) shade that cylindrical object is formed is certain
It is axisymmetric, and its axis of symmetry is bound to through the center of circle of the circular cross section;(2) in two-dimensional space, two differences
Point can uniquely determine straight line.
Step 5:The axis of symmetry and shade thickness obtained according to step 4, the stocking rate of flat-top tank is calculated;
Step 6:Because true reserve data are difficult to obtain, based on vision offset caused by shooting angle, manually
Reservoir data is calculated, is mutually authenticated with real data.
Described vision offset refers to the presence due to shooting angle, edge, the edge and tank body of floating roof at the top of tank body
The edge of bottom can stagger on two dimensional surface.
The specific implementation process of above-mentioned each module is as follows:
1. the candidate target extraction based on improved ELSD algorithms
Implementation process is as shown in Fig. 2 be divided into following 4 sub-steps:
1) verify whether two discrete circular arcs belong to same annulus, that is, determine the membership of circular arc.It is main logical
Following three formula are crossed to complete:
Here Cenr, Cenc represent the row, column coordinate in the center of circle of some circular arc respectively, and r represents the radius of circular arc.Subscript a,
B represents two different circular arcs.θ, σ are small constant, represent specific threshold value.Formula (2) is used to whether confirm two circular arcs
With the same center of circle;Formula (3) is used to confirm whether two circular arcs have identical radius.
2) number of pixels for all circular arc fragments corresponding to same annulus that calculating detects, which accounts for the annulus, should pixel
The ratio R of numbercircle, as shown in formula (4):
Here lcircleRepresent all pixels points on a certain annulus, kx(circle) represent to detect on corresponding annulus
Circular arc total pixel number.It can obtain, RcircleValue between 0 to 1, and its value is closer to 1, respective regions
More it is likely to be target.Given threshold R0If required RcircleMore than R0, then retain the candidate region, otherwise delete the candidate regions
Domain.
3) candidate region is grouped according to Jaccard similarities, it is i.e. corresponding equivalent to space clustering, each packet
In unique target.
Here area represents the area of some circle detected, and subscript c, d represents different candidate regions, when calculating
In order to simple, directly with circular external square cartographic represenation of area.
4) for each packet, then according to space length two class clusters of progress, it is therefore an objective to determined most for each target
Unique candidate region eventually.The present invention selects K-means clustering algorithms.
Each cluster the mean circle-center coordinate of all candidate regions and radius corresponding to it represent, i.e., form for (Cenr,
Cenc, r) triple.If the triple of two clusters meets formula (2), formula (3) is but unsatisfactory for, illustrates two clusters
The center of circle is consistent, and radius is inconsistent, then the triple of the larger cluster of radius is using as the spatial information of respective objects.It is because this
In the case of, the cluster with shorter radius is likely to as caused by the inside circular profile on flat-top tank top, and radius is smaller
Cluster should be rejected.In addition, i.e., the center of circle is inconsistent, then quantity less cluster in candidate region is likely to be produced by shade
Raw.Now the triple of the larger cluster of radix is using as the spatial information of respective objects.
2. train and optimize DCNN models
DCNN network structures used in the present invention are as shown in figure 3, the network includes three convolutional layers
(Convolutional layer), three pond layers (Pooling layer), a full articulamentum (Inner product
Layer) and a softmax layer, wherein pond layer include maximum pond layer (Max pooling layer) and average pond layer
(Average pooling layer).DCNN increases income in Caffe and realized on deep learning framework in embodiment.Activation primitive makes
It is a kind of nonlinear activation function with amendment linear unit (Rectified Linear Units, ReLU).Network weight
(Weights) the Gaussian Profile initialization that variance is 0 for 0.01, average is used;Bias (Bias) is directly initialized as 0;Study
Rate (Learning rate) is initialized as 0.001, afterwards as the increase of iteration wheel number progressively decays.
3. based on the shade symmetry analysis method put to comparing
Implementation process is as shown in figure 4, be divided into following 4 sub-steps:
1) center of circle of tank body cross section is set to fixing point, travels through institute on circular arc a little, point and circle on each circular arc
The heart all uniquely determines straight line;
Whether following two formula are used to verify a point in circular arc:
Here, Cen and P represents any point on the center of circle and circular arc respectively, and r represents the radius of circular arc.Subscript represents one
The row, column coordinate of individual point.θ is a small constant, represents given threshold, and the present invention takes 0.1, can control the thickness of circular arc.
Formula (6) is used to calculate point the distance between Cen and P;Formula (7) is used for whether check post P to define in center of circle Cen and radius r
Circular arc on.
2) for each point in circle, verify whether it is consistent with its on the pixel value of the 1) symmetric points of gained straight line;
For check post Q in the circle that Cen and r are defined, formula is as follows:
Here Cen and r definition is same as above, and Q represents any point in image.σ is a small constant, is represented
Given threshold, the present invention take 1.1.
Following formula is used to calculate certain symmetric points of point on straight line.Linear equation is with the standard of formula (9)
Form represents.
Ax+By+C=0 (9)
x1=x0-2×A×T (11)
y1=y0-2×B×T (12)
Here A, B, C represent the coefficient of linear equation.(x0,y0) represent circle in any point;(x1,y1) represent (x0,
y0) symmetric points on the straight line.
3) for every straight line, the number of the interior inconsistent point pair on the straight line of statistics circle;
4) according to the inconsistent minimum straight line of logarithm 3), is obtained, exported as the final axis of symmetry.
Here PfinalOne in two points that the symmetry axis finally calculated intersects with circular arc is represented, it is unique with the center of circle
The axis of symmetry is determined.PiRepresent a bit in circle, Pi' represent PiSymmetric points on symmetry axis.Pi=Pi' represent two
Point is all shade or is not.Formula (13) finds out an intersection point of the inconsistent minimum straight line of logarithm and circular arc;Formula
(14) it is the indicator function that represents value condition, for constructing statistic.
4. the stocking rate computational methods based on shade thickness
As shown in figure 5, flat-top tank in top float, can form two parts shade.Shade on the inside of circular arc is by tank
What the difference in height of body sidewall and flat-top was formed;Shade on the outside of circular arc is formed by the difference in height on tank wall and ground.
The l that they are corresponded respectively in Fig. 5inAnd lout.It is divided into following 3 sub-steps:
1) intersecting point coordinate of symmetry axis and circular arc is found, and calculates the shadows pixels on the axis of symmetry between two intersection points
Count out, as partial phantom thickness lin;
2) incident direction of light is determined according to the offset direction of shade in 1), and determined according to incident direction outside tank body
The position of shade, calculate the length l that backlight side dash area intersects with symmetry axisout;
According to parallel projection principle and triangle correspondence theorem, 1) and 2) 3) ratio of two parts shade thicknessIt is equal to
The vacancy rate of oil tankAccordingly obtain the stocking rate of oil tank
Vacancy rate calculation formula is as follows:
Stocking rate calculation formula is as follows:
Here rreserveRepresent the stocking rate of oil tank, linAnd loutThe shade of tank body cross section inner side and outer side is represented respectively
The length that part is intersected with symmetry axis.h1And h2Represent tank deck edge relative to the difference in height of ground level and flat-top, such as Fig. 5 respectively
Shown in middle darker regions.
5. the true value method of estimation of view-based access control model skew
Due to the presence of shooting angle, the edge at edge, floating roof at the top of tank body and the edge of tank base can be in two dimensions
Stagger in plane.As shown in fig. 6, the center of circle for the three parts annulus that gray line marks is offset in shooting direction along straight line.
Fig. 7 illustrates the reason for this phenomenon formation.From the angle of surface, the center of circle of three parts should overlap one to hang down
On straight line.But when shooting angle secundly, two layers above is visually offset.Along the straight line of shooting angle
Direction, with respect to bottom, top layer offset by d1Distance;With respect to intermediate layer, top layer offset by d2Distance.Meanwhile top layer relative to
The difference in height of bottom is h1, the difference in height relative to intermediate layer is h2.Arrow in Fig. 7 beside " visual angle " is view directions.
1) as shown in fig. 7, artificial measurement obtains corresponding shade thickness d1、d2;
2) according to triangle correspondence theorem, they meet the proportionate relationship of formula (17), obtain vacancy ratePublic affairs are substituted into again
Formula (16), obtain another stocking rate dataAs experimental result true value, for carrying out experimental evaluation compared with;
3) experimental result is compared with the true value 2) being calculated, and using root-mean-square error as evaluation criterion, error is got over
It is small, represent that experimental result is more reliable.
Here rreserveRepresent experimental result, rgroundArtificial result of calculation is represented, m represents sample size.Err represents equal
Square error, represent the similarity degree between data.
Non-elaborated part of the present invention belongs to techniques well known.
It is described above, part embodiment only of the present invention, but protection scope of the present invention is not limited thereto, and is appointed
What those skilled in the art the invention discloses technical scope in, it will be appreciated that the change or replacement expected, should all cover
Within protection scope of the present invention.Therefore, protection scope of the present invention should be defined by the protection domain of claims.
Claims (5)
1. oil tank detection and reserve analysis method based on high-resolution optical remote sensing image, it is characterised in that including following step
Suddenly:
Step 1:Gather and mark the remote sensing image data of the target containing oil tank, split data into training set, checking collection and survey
Examination collection;
Step 2:Based on improved ELSD (Ellipse and Line Segment Detector) algorithm, step 1 is obtained
Training set, checking collection and test set extract training set, checking collection and test set candidate target respectively;
Step 3:Training set, the checking collection candidate target obtained using step 2 is trained and optimizes DCNN (Deep
Convolutional Neural Network) model, the DCNN that the test set candidate target input that step 2 obtains has been trained
Model, exports target classification result, and the target classification result includes flat-top tank, arch-roof tank and the class of background three;
Step 4:Using the symmetrical feature of flat-top tank surrounding shadow, based on point to the shade symmetry analysis method compared, extract flat
Push up the optimal axis of symmetry in tank surrounding shadow region;
Step 5:The axis of symmetry obtained according to shade thickness and step 4, the stocking rate of flat-top tank is calculated;
Step 6:True reserve data are difficult to obtain, and based on vision offset caused by shooting angle, calculate deposit manually
Data, it is mutually authenticated with real data.
2. oil tank detection and reserve analysis method according to claim 1 based on high-resolution optical remote sensing image, its
It is characterised by:The specific implementation of the step 2 includes following sub-step:
Step 2.1:Whether two discrete circular arcs of checking belong to same annulus, that is, determine the membership of circular arc, by with
Lower three formula provide:
<mrow>
<mi>D</mi>
<mi>i</mi>
<mi>s</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mo>,</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msqrt>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>Cenr</mi>
<mi>a</mi>
</msub>
<mo>-</mo>
<msub>
<mi>Cenr</mi>
<mi>b</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>Cenc</mi>
<mi>a</mi>
</msub>
<mo>-</mo>
<msub>
<mi>Cenc</mi>
<mi>b</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mfrac>
<mrow>
<mi>D</mi>
<mi>i</mi>
<mi>s</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mo>,</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>r</mi>
<mi>a</mi>
</msub>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>b</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>&le;</mo>
<mi>&theta;</mi>
<mo>,</mo>
<mi>&theta;</mi>
<mo>&GreaterEqual;</mo>
<mn>0</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mfrac>
<mrow>
<mo>|</mo>
<mrow>
<msub>
<mi>r</mi>
<mi>a</mi>
</msub>
<mo>-</mo>
<msub>
<mi>r</mi>
<mi>b</mi>
</msub>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>r</mi>
<mi>a</mi>
</msub>
<mo>,</mo>
<msub>
<mi>r</mi>
<mi>b</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>&le;</mo>
<mi>&sigma;</mi>
<mo>,</mo>
<mi>&sigma;</mi>
<mo>&GreaterEqual;</mo>
<mn>0</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Cenr, Cenc represent the row, column coordinate in the center of circle of some circular arc respectively, and r represents the radius of circular arc, and subscript a, b represents two
Individual different circular arc, θ, σ are small constant, represent the threshold value of setting, and formula (2) is used to confirm whether two circular arcs have equally
The center of circle;Formula (3) is used to confirm whether two circular arcs have identical radius;
Step 2.2:The number of pixels for all circular arc fragments corresponding to same annulus that calculating detects, which accounts for the annulus, should picture
Prime number purpose ratio;
<mrow>
<msub>
<mi>R</mi>
<mrow>
<mi>c</mi>
<mi>i</mi>
<mi>r</mi>
<mi>c</mi>
<mi>l</mi>
<mi>e</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>k</mi>
<mi>x</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>c</mi>
<mi>i</mi>
<mi>r</mi>
<mi>c</mi>
<mi>l</mi>
<mi>e</mi>
<mo>)</mo>
</mrow>
</mrow>
<msub>
<mi>l</mi>
<mrow>
<mi>c</mi>
<mi>i</mi>
<mi>r</mi>
<mi>c</mi>
<mi>l</mi>
<mi>e</mi>
</mrow>
</msub>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
lcircleRepresent all pixels points on a certain annulus, kx(circle) circular arc detected on the corresponding annulus of expression is total
Pixel number, RcircleValue between 0 to 1, and its value, closer to 1, respective regions are more likely to be target;Setting
Certain threshold value R0, the candidate region after being filtered;
Step 2.3:It is grouped according to the candidate region that Jaccard similarities obtain step 2.2, equivalent to space clustering,
Each packet corresponds to unique target;
<mrow>
<mi>J</mi>
<mrow>
<mo>(</mo>
<mi>c</mi>
<mo>,</mo>
<mi>d</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<mrow>
<msub>
<mi>area</mi>
<mi>c</mi>
</msub>
<mo>&cap;</mo>
<msub>
<mi>area</mi>
<mi>d</mi>
</msub>
</mrow>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<mrow>
<msub>
<mi>area</mi>
<mi>c</mi>
</msub>
<mo>&cup;</mo>
<msub>
<mi>area</mi>
<mi>d</mi>
</msub>
</mrow>
<mo>|</mo>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
Area represents the area of some circle detected, and subscript c, d represents the different candidate regions that step 2.2 obtains,
In order to simple during calculating, directly with circular external square cartographic represenation of area;
Step 2.4:Two class clusters are carried out for each packet, then according to space length, it is final only for the determination of each target
One candidate region, so far extraction obtain candidate target;
Each cluster the mean circle-center coordinate of all candidate regions and radius corresponding to it represent, i.e., form for (Cenr,
Cenc, r) triple, if two cluster triples meet formula (2), be but unsatisfactory for formula (3), illustrate two cluster
The center of circle is consistent, and radius is inconsistent, then the triple of the larger cluster of radius is using as the spatial information of respective objects;It is because this
In the case of, the cluster with shorter radius is likely to as caused by the inside circular profile on flat-top tank top, and radius is smaller
Cluster should be rejected;In addition, i.e. the center of circle is inconsistent, then quantity less cluster in candidate region is likely to be produced by shade
, now the triple of the larger cluster of radix is using as the spatial information of respective objects.
3. oil tank detection and reserve analysis method according to claim 1 based on high-resolution optical remote sensing image, its
It is characterised by:The specific implementation of step 4 includes following sub-step:
Step 4.1:The center of circle of tank body cross section is set to fixing point, travels through institute on circular arc a little, point on each circular arc and
The center of circle all uniquely determines straight line;
Step 4.2:Using the symmetrical feature of flat-top tank surrounding shadow, for each point in circle, verify each point on step
Whether the pixel value of the symmetric points of 4.1 gained straight lines is consistent with each pixel value;
Step 4.3:For every straight line, the number of the interior inconsistent logarithm on the straight line of statistics circle;
Step 4.4:According to step 4.3, the minimum straight line of inconsistent logarithm is obtained, is exported as the final axis of symmetry.
4. oil tank detection and reserve analysis method according to claim 1 based on high-resolution optical remote sensing image, its
It is characterised by:The specific implementation of step 5 includes following sub-step:
Step 5.1:The intersecting point coordinate of the axis of symmetry obtained in step 4 and circular arc is found, and calculates pair between two intersection points
The shadows pixels on axis are claimed to count out, as shade thickness lin;
Step 5.2:The incident direction of light is determined according to the offset direction of shade in step 5.1, and is determined according to incident direction
The position of shade outside tank body, calculate the shade thickness l that backlight side dash area intersects with symmetry axisout;
Step 5.3:According to parallel projection principle and triangle correspondence theorem, step 5.1 and step 5.2 two parts shade thickness
RatioEqual to the vacancy rate of oil tankSo as to obtain the stocking rate of flat-top tankWherein h1And h2Tank is represented respectively
Top is relative to ground and the difference in height of flat-top.
5. oil tank detection and reserve analysis method according to claim 1 based on high-resolution optical remote sensing image, its
It is characterised by:The specific implementation of step 6 includes following sub-step:
Step 6.1:Due to the presence of shooting angle, the edge at edge, floating roof at the top of tank body and the edge of tank base are two
Stagger on dimensional plane, artificial measurement obtains corresponding shade thickness d1、d2;
Step 6.2:According to triangle correspondence theorem, the shade thickness d in step 6.11、d2RatioIt is vacant equal to oil tank
RateObtain another stocking rate dataAs experimental result true value, for carrying out experimental evaluation compared with;
Step 6.3:When experimental result compares with the true value that step 6.2 is calculated, marked using root-mean-square error as evaluation
Standard, error is smaller, represents that experimental result is more reliable;
<mrow>
<mi>e</mi>
<mi>r</mi>
<mi>r</mi>
<mo>=</mo>
<msqrt>
<mrow>
<mfrac>
<mn>1</mn>
<mi>m</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>r</mi>
<mrow>
<mi>r</mi>
<mi>e</mi>
<mi>s</mi>
<mi>e</mi>
<mi>r</mi>
<mi>v</mi>
<mi>e</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>r</mi>
<mrow>
<mi>g</mi>
<mi>r</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>d</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
rreserveRepresent experimental result, rgroundArtificial result of calculation is represented, m represents sample size, and err represents root-mean-square error,
Represent the similarity degree between data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710727161.4A CN107688782A (en) | 2017-08-23 | 2017-08-23 | Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710727161.4A CN107688782A (en) | 2017-08-23 | 2017-08-23 | Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107688782A true CN107688782A (en) | 2018-02-13 |
Family
ID=61153605
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710727161.4A Pending CN107688782A (en) | 2017-08-23 | 2017-08-23 | Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107688782A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110210453A (en) * | 2019-06-14 | 2019-09-06 | 中国资源卫星应用中心 | A kind of oil tank amount of storage based on Characteristics of The Remote Sensing Images determines method and system |
CN111462221A (en) * | 2020-04-03 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Method, device and equipment for extracting shadow area of object to be detected and storage medium |
CN111462222A (en) * | 2020-04-03 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Method, device, equipment and medium for determining reserve of object to be detected |
CN111540007A (en) * | 2020-04-10 | 2020-08-14 | 中国资源卫星应用中心 | Method for estimating storage capacity of oil tank by utilizing SAR (synthetic aperture radar) image |
CN112200858A (en) * | 2020-10-10 | 2021-01-08 | 长光卫星技术有限公司 | External floating roof oil tank reserve analysis method based on high-resolution optical remote sensing image |
CN112749673A (en) * | 2021-01-20 | 2021-05-04 | 西安中科星图空间数据技术有限公司 | Method and device for intelligently extracting stock of oil storage tank based on remote sensing image |
CN113160173A (en) * | 2021-04-22 | 2021-07-23 | 哈尔滨市科佳通用机电股份有限公司 | Oil leakage detection method and system of snake-shaped-resistant shock absorber based on Laws texture features |
CN113592033A (en) * | 2021-08-20 | 2021-11-02 | 中科星睿科技(北京)有限公司 | Oil tank image recognition model training method, oil tank image recognition method and oil tank image recognition device |
CN113781478A (en) * | 2021-11-09 | 2021-12-10 | 中科星睿科技(北京)有限公司 | Oil tank image detection method, oil tank image detection device, electronic equipment and computer readable medium |
CN113808134A (en) * | 2021-11-19 | 2021-12-17 | 中科星睿科技(北京)有限公司 | Oil tank layout information generation method, oil tank layout information generation device, electronic apparatus, and medium |
CN115984836A (en) * | 2023-03-20 | 2023-04-18 | 山东杨嘉汽车制造有限公司 | Tank opening identification and positioning method for railway tank wagon |
CN111462222B (en) * | 2020-04-03 | 2024-05-14 | 深圳前海微众银行股份有限公司 | Method, device, equipment and medium for determining reserves of objects to be detected |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1373349A (en) * | 2002-04-02 | 2002-10-09 | 上海交通大学 | Oil tank volume measuring system and calibrating method |
CN104217196A (en) * | 2014-08-26 | 2014-12-17 | 武汉大学 | A method for detecting automatically a circular oil tank with a remote sensing image |
CN106909902A (en) * | 2017-03-01 | 2017-06-30 | 北京航空航天大学 | A kind of remote sensing target detection method based on the notable model of improved stratification |
-
2017
- 2017-08-23 CN CN201710727161.4A patent/CN107688782A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1373349A (en) * | 2002-04-02 | 2002-10-09 | 上海交通大学 | Oil tank volume measuring system and calibrating method |
CN104217196A (en) * | 2014-08-26 | 2014-12-17 | 武汉大学 | A method for detecting automatically a circular oil tank with a remote sensing image |
CN106909902A (en) * | 2017-03-01 | 2017-06-30 | 北京航空航天大学 | A kind of remote sensing target detection method based on the notable model of improved stratification |
Non-Patent Citations (2)
Title |
---|
QINGQUAN WANG 等: "Automatic Detection and Classification of Oil Tanks in Optical Satellite Images Based on Convolutional Neural Network", 《INTERNATIONAL CONFERENCE ON IMAGE AND SIGNAL PROCESSING》 * |
VERONICA MAGAN: "Orbital Insight Measures China Oil Supply With Satellite Imagery Analysis", 《HTTPS://WWW.SATELLITETODAY.COM/INNOVATION/2016/09/30/ORBITAL-INSIGHT-MEASURES-CHINA-OIL-SUPPLY-SATELLITE-IMAGERY-ANALYSIS/》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110210453A (en) * | 2019-06-14 | 2019-09-06 | 中国资源卫星应用中心 | A kind of oil tank amount of storage based on Characteristics of The Remote Sensing Images determines method and system |
CN110210453B (en) * | 2019-06-14 | 2021-06-29 | 中国资源卫星应用中心 | Remote sensing image feature-based oil tank storage capacity determination method and system |
WO2021196698A1 (en) * | 2020-04-03 | 2021-10-07 | 深圳前海微众银行股份有限公司 | Method, apparatus and device for determining reserve of object to be detected, and medium |
CN111462221A (en) * | 2020-04-03 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Method, device and equipment for extracting shadow area of object to be detected and storage medium |
CN111462222A (en) * | 2020-04-03 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Method, device, equipment and medium for determining reserve of object to be detected |
CN111462222B (en) * | 2020-04-03 | 2024-05-14 | 深圳前海微众银行股份有限公司 | Method, device, equipment and medium for determining reserves of objects to be detected |
CN111540007A (en) * | 2020-04-10 | 2020-08-14 | 中国资源卫星应用中心 | Method for estimating storage capacity of oil tank by utilizing SAR (synthetic aperture radar) image |
CN111540007B (en) * | 2020-04-10 | 2023-04-28 | 中国资源卫星应用中心 | Method for estimating oil tank storage capacity by SAR image |
CN112200858A (en) * | 2020-10-10 | 2021-01-08 | 长光卫星技术有限公司 | External floating roof oil tank reserve analysis method based on high-resolution optical remote sensing image |
CN112749673A (en) * | 2021-01-20 | 2021-05-04 | 西安中科星图空间数据技术有限公司 | Method and device for intelligently extracting stock of oil storage tank based on remote sensing image |
CN113160173A (en) * | 2021-04-22 | 2021-07-23 | 哈尔滨市科佳通用机电股份有限公司 | Oil leakage detection method and system of snake-shaped-resistant shock absorber based on Laws texture features |
CN113592033A (en) * | 2021-08-20 | 2021-11-02 | 中科星睿科技(北京)有限公司 | Oil tank image recognition model training method, oil tank image recognition method and oil tank image recognition device |
CN113592033B (en) * | 2021-08-20 | 2023-09-12 | 中科星睿科技(北京)有限公司 | Oil tank image recognition model training method, oil tank image recognition method and device |
CN113781478A (en) * | 2021-11-09 | 2021-12-10 | 中科星睿科技(北京)有限公司 | Oil tank image detection method, oil tank image detection device, electronic equipment and computer readable medium |
CN113781478B (en) * | 2021-11-09 | 2022-05-24 | 中科星睿科技(北京)有限公司 | Oil tank image detection method, oil tank image detection device, electronic equipment and computer readable medium |
CN113808134A (en) * | 2021-11-19 | 2021-12-17 | 中科星睿科技(北京)有限公司 | Oil tank layout information generation method, oil tank layout information generation device, electronic apparatus, and medium |
CN115984836A (en) * | 2023-03-20 | 2023-04-18 | 山东杨嘉汽车制造有限公司 | Tank opening identification and positioning method for railway tank wagon |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107688782A (en) | Oil tank detection and reserve analysis method based on high-resolution optical remote sensing image | |
Huang et al. | A multidirectional and multiscale morphological index for automatic building extraction from multispectral GeoEye-1 imagery | |
CN110287932B (en) | Road blocking information extraction method based on deep learning image semantic segmentation | |
Sirmacek et al. | Urban-area and building detection using SIFT keypoints and graph theory | |
Witharana et al. | Optimizing multi-resolution segmentation scale using empirical methods: Exploring the sensitivity of the supervised discrepancy measure Euclidean distance 2 (ED2) | |
Turker et al. | Field-based sub-boundary extraction from remote sensing imagery using perceptual grouping | |
Awrangjeb et al. | Building detection in complex scenes thorough effective separation of buildings from trees | |
Chaudhuri et al. | Semi-automated road detection from high resolution satellite images by directional morphological enhancement and segmentation techniques | |
CN104778721B (en) | The distance measurement method of conspicuousness target in a kind of binocular image | |
CN107330376A (en) | A kind of Lane detection method and system | |
CN108830188A (en) | Vehicle checking method based on deep learning | |
San et al. | Building extraction from high resolution satellite images using Hough transform | |
CN106339674A (en) | Hyperspectral image classification method based on edge preservation and graph cut model | |
CN106558072A (en) | A kind of method based on SIFT feature registration on remote sensing images is improved | |
CN104657717B (en) | A kind of pedestrian detection method based on layering nuclear sparse expression | |
Yuan et al. | Learning to count buildings in diverse aerial scenes | |
CN103034863A (en) | Remote-sensing image road acquisition method combined with kernel Fisher and multi-scale extraction | |
CN102945374B (en) | Method for automatically detecting civil aircraft in high-resolution remote sensing image | |
CN103761526B (en) | Urban area detecting method based on feature position optimization and integration | |
CN104182985A (en) | Remote sensing image change detection method | |
CN102629380B (en) | Remote sensing image change detection method based on multi-group filtering and dimension reduction | |
Peng et al. | Object-based change detection from satellite imagery by segmentation optimization and multi-features fusion | |
Zang et al. | Road network extraction via aperiodic directional structure measurement | |
CN113963222A (en) | High-resolution remote sensing image change detection method based on multi-strategy combination | |
CN103366373A (en) | Multi-time-phase remote-sensing image change detection method based on fuzzy compatible chart |
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 |
Application publication date: 20180213 |
|
WD01 | Invention patent application deemed withdrawn after publication |