CN103150582B - Modified result method is sentenced based on contextual remote sensing image cloud - Google Patents
Modified result method is sentenced based on contextual remote sensing image cloud Download PDFInfo
- Publication number
- CN103150582B CN103150582B CN201310043710.8A CN201310043710A CN103150582B CN 103150582 B CN103150582 B CN 103150582B CN 201310043710 A CN201310043710 A CN 201310043710A CN 103150582 B CN103150582 B CN 103150582B
- Authority
- CN
- China
- Prior art keywords
- block
- cloud
- decision value
- atural object
- piecemeal
- 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.)
- Active
Links
Abstract
The invention discloses one and sentence modified result method based on contextual remote sensing image cloud, do not need to set up probability model, sentence problem design for cloud, utilize the court verdict of adjacent piecemeal and background information to carry out revised version block court verdict simultaneously.The first step, reads in pending related data: comprise the block index matrix of former figure, piecemeal gray average matrix and background information; Second step, utilizes the decision value of Tile to produce the decision value of Block, forms Block index matrix, and set up Block gray average matrix according to piecemeal gray average matrix; 3rd step, utilizes decision value to be the Block determining very much cloud and determine very much atural object, and revising decision value is uncertain Block decision value; 4th step, the false-alarm of the background of application target context and scene context determination handling object, the false-alarm rejecting atural object intersection, rejecting cloudlet also adjusts the adjacent target of the accurate object of judgement, and whether finally determine Block is cloud.
Description
Technical field
The present invention relates to one and sentence modified result method based on contextual remote sensing image cloud, belong to mode identification technology.
Background technology
Model most widely used in current computer vision is the theory of vision computing of Mary.Although it has greatly promoted the development of computer vision research, be considered to perfect not enough to a great extent, mainly because the selectivity do not comprised in human vision and globality.So-called selectivity refers to top-down processing mode under contextual guidance, and globality just refers to the strategy of global precedence effect.
There is a lot of research to use for reference some viewpoints of human vision at present, proposed based on contextual object detection and recognition method.
Wherein, context can be defined as relevant with target but not be all information of the apparent description of target itself.Context can be divided into following three different levels by us: local context (characteristic layer), target context (destination layer) and scene context (scene layer).
1) local context
The object detection and recognition method of current main flow based on local feature because it have to simple, easily calculate, to advantages such as affined transformation etc. are insensitive, in addition for blocking, change in illumination and class and also have certain repellence.But it does not consider the position relationship between local feature, and this relation is very important for object detection and recognition.
2) target context
Target context refers to cooccurrence relation and the position relationship of target to be identified and other object in scene.Certainly, the detection and indentification utilizing other object in scene to treat recognition target object is very helpful.But based target context approach but needs a condition, namely the identification of other object is more accurately.
3) scene context
The mankind from entirety, so have some research methods recently from the feature of whole scene, carry out scene classification task, and achieve extraordinary effect in scene Recognition.But these methods also comprise some intermediate steps, such as segmentation, feature organization and target identification, and these problems are also the key issues solved not yet completely at present in computer vision.
In sum, the context of above three levels is utilized, various different emerging in an endless stream based on contextual target identification theory.Such as, based on the word bag (Bag-of-features of local context, BOF) model, Markov random field (MakovRandomField, MRF), discriminative random fields (DiscriminativeRandomFields, DRF); What based target context Fink etc. proposed utilizes the target detection structure of cascade to detect local feature and other order calibration method of target; A kind of super characteristic bag (BeyondBag-of-Features, BBoFs) model is proposed based on scene context Lazebnik etc.Mostly there is a common feature in these methods, needs exactly to set up probability model.First, this adds increased the difficulty of this algorithm in hard-wired; Secondly, spatial-domain information very important in image is not made full use of yet; Finally, these are all universal algorithms, do not sentence the singularity of problem for cloud.
So, can consider directly to start with from the logic feature of cloud and atural object space distribution, main research, according to the local context of the pattern recognition result of space distribution and scene context, finally obtains a kind of fully shelf space information and is applicable to hard-wired context correction technique.
Summary of the invention
For the problems referred to above, the invention provides one and sentence modified result method based on contextual remote sensing image cloud, do not need to set up probability model, but directly start with from the logic feature of cloud and atural object space distribution, sentence problem design for cloud, utilize the court verdict of adjacent piecemeal and background information to carry out revised version block court verdict simultaneously.It differentiates that accuracy increases, and realizes relatively simple.
Modified result method should be sentenced based on contextual remote sensing image cloud, comprise the following steps:
The first step, reads in pending related data: comprise the block index matrix of former figure, piecemeal gray average matrix and background information;
Described piecemeal is the stepping adopting the square of L × L to carry out level and vertical direction in former figure, stepping-in amount is L/2, obtain multiple piecemeal, be called Tile, every four adjacent overlapped Tile piecemeal can regard nine nonoverlapping sizes as is up and down that the little piecemeal of L/2 × L/2 is formed, and this little piecemeal is called Block;
Element in described block index matrix is the pattern-recognition decision value of each Tile piecemeal, comprises cloud, atural object and uncertain;
Element in described piecemeal gray average matrix is the gray average of each Tile piecemeal;
Described background information comprises atural object thresholding and Yunmen limit, and both is all the global threshold of corresponding former figure;
Second step, the context that application overlap partition produces, specifically comprises step S21 ~ S22;
Step S21, utilizes the decision value of Tile to produce the decision value of Block, forms Block index matrix; Divide three kinds of situations;
When Block is in the position at four angles, obtain the decision value of unique Tile belonging to it, if this decision value is cloud or atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain;
When Block is in the position on four limits: the decision value obtaining its affiliated two Tile, if two decision values are all clouds or are all atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain;
When Block is in other positions: the decision value obtaining its affiliated four Tile, when having a cloud and an atural object in four decision values at least, then the decision value of Block is uncertain; When the number being judged to cloud in four decision values is four, then the decision value of Block is for very to determine cloud; When the number being judged to atural object in four decision values is four, then the decision value of Block is for very to determine atural object; When only comprise cloud in four decision values and uncertain time, then the decision value of Block is more accurate cloud; When only comprise atural object in four decision values and uncertain time, then the decision value of Block is more accurate atural object;
Step S22, sets up Block gray average matrix according to piecemeal gray average matrix, the corresponding each Block of its element, and element value is the gray average of Block, adopts the gray average of each Tile belonging to the Block that extracts from piecemeal gray average matrix to be averaged again and obtains;
3rd step, utilizes decision value to be the Block determining very much cloud and determine very much atural object, and revising decision value is uncertain Block decision value:
The Block with eight neighborhoods is traveled through successively by order from left to right, from top to bottom, if the decision value of Block is for very determining cloud or determining very much atural object, then the court verdict each Block on its eight neighborhood being judged to the Block of the more accurate type of corresponding class becomes to be determined very much;
4th step, application target context and scene context, specifically comprise step S41 ~ S45:
Step S41, judges the background of handling object: when the atural object thresholding in background information is less than the sea route boundary thresholding of setting, background thinks sea, otherwise is land; Corresponding background is the situation in sea, reduces Yunmen limit, otherwise, raise Yunmen limit; The modification rule of Yunmen limit is artificially determined;
Step S42, reject the false-alarm of atural object intersection: in the Block of 3 × 3 sizes, the decision value of central authorities Block is thing very definitely, and to there is adjacent Block decision value be uncertain, more accurate cloud or more accurate atural object, and its gray average is not more than central Block's, then the decision value of this adjacent Block is modified to and very determines atural object;
Step S43, reject the false-alarm of cloudlet: in the Block of 3 × 3 sizes, if the decision value of central Block is uncertain, more accurate cloud or more accurate atural object, and eight Block major parts around are all judged to atural object, and the gray average of central Block be greater than one setting cloud gray scale thresholding time, then the decision value of central Block is modified to and very determines cloud;
Step S44, the adjacent target of the accurate object of adjustment judgement: in the Block of 3 × 3 sizes, if the decision value of central Block is for very to determine cloud, then adjacent uncertain Block is corrected for more accurate atural object, in like manner, if the decision value of central Block is for very to determine atural object, then adjacent uncertain Block is corrected for more accurate cloud;
Step S45, the process of scene context is utilized to be left target: to the Block being still judged to uncertain, more accurate cloud or more accurate atural object through step above, the Yunmen obtained in the gray average of this Block and step S41 limit is compared, if be greater than, then this Block is judged to and very determines cloud, otherwise is judged to and very determines atural object; So far whole process terminates.
Beneficial effect:
Local context, target context and background context considered in context of the present invention, directly starts with from the logic feature of cloud and atural object space distribution, sentence problem design for cloud, differentiates that accuracy increases.And the present invention does not set up probability model, differentiate that process computation is simple, this reduces the difficulty of this algorithm in hard-wired.
Accompanying drawing explanation
Fig. 1 sentences modified result method flow diagram for the present invention is based on contextual remote sensing image cloud;
Fig. 2 is piecemeal operation chart;
Fig. 3 is the process flow diagram of second step in Fig. 1;
Fig. 4 is the process flow diagram of the 4th step in Fig. 1.
Embodiment
The present invention is the context correction technique carried out based on index matrix, so-called index matrix, refers to that Land use models recognition methods judgement obtains classification results, and these results are formed a matrix according to specific rule.Some numerals of usual classification results represent different classes, such as, can represent be judged to cloud with positive number, negative number representation is judged to atural object, represent the size of the degree of confidence being judged to such by the size of the absolute value of these numbers, when it equals zero, illustrate that court verdict is uncertain.So the data volume handled by the present invention utilizes the recognition methods of former figure much smaller relative to needing, this advantage merits attention in the hardware realizability of algorithm.
To develop simultaneously embodiment below in conjunction with accompanying drawing, describe the present invention.
The first step, read in pending related data:
Modification method of the present invention revises for the result of a former figure after pattern recognition process.Wherein, read in pending related data and comprise the index matrixs of composition such as pattern recognition result and the background information of this width figure.Index matrix refers to the decision value matrix of the decision value composition of the pattern-recognition of the corresponding each Tile piecemeal of former figure and the gray average matrix of gray average composition, and background information refers to the atural object thresholding that former figure is corresponding and Yunmen limit.Meanwhile, pattern-recognition needs to adopt specific partitioned mode, describes in detail below to partitioned mode:
As shown in Figure 2, first, choosing according to whole figure the block size size will carrying out pattern recognition process is L × L, and making this piecemeal be Tile, is L/2 with seasonal level and vertical stepping.Then, by this stepping-in amount, piecemeal is carried out to each row from left to right, finally, moved the piecemeal to full figure according to this stepping more from top to bottom with the partitioned mode of this row.So can draw, adjacent four the Tile piecemeals (distribution of similar field word) overlapped each other, also can regard the little piecemeal formation that nine nonoverlapping sizes are L/2 × L/2 as, make this little piecemeal be Block.Further, Block is overlapping minimum unit, and each Tile can be divided into four Block, most Block all comprise by around four different Ti le.
Based on above-mentioned piecemeal, the element in described block index matrix is the pattern-recognition decision value of each Tile piecemeal, comprises cloud, atural object and uncertain; Element in described piecemeal gray average matrix is the gray average of each Tile piecemeal.
The context that second step, application overlap partition produce:
According to the index matrix that the pattern recognition result of above-mentioned overlap partition forms, the inside lap (namely Block) and surrounding is utilized to contain the particular kind of relationship of its piecemeal (namely Tile), the pattern recognition result of these piecemeals is comprehensively become the court verdict of corresponding lap by certain rule, court verdict can be divided into very accurately accuracy, more accurate and uncertain, can be divided into algorithm object cloud, atural object and both mixing.Thus improve recognition resolution, follow-up process be context relation for lap.
Concrete steps in this step:
Step S21, the decision value of Block is determined: due to the method for partition in the first step according to block index matrix, whole figure also can regard as and form by Block is closely non-overlapping, and like this when the position of Block is in the position at four angles in whole figure, Block can only be comprised in a Tile; When being in the position of four edges (not comprising angle) in whole figure, Block can only be comprised in two Tile; In addition other positions Block all comprise by around four different Ti le.So all Block are divided three classes according to above-mentioned position relationship by we:
(1) when Block is in the position at four angles: because only have a court verdict, so directly inherit original court verdict.Obtain the decision value of unique Tile belonging to it, if this decision value is cloud or atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain.
(2) when Block is in the position of four limits (not comprising angle): now Block is to there being two court verdicts, obtain the decision value of its affiliated two Tile, if two decision values are all clouds or are all atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain.
(3), when Block is in other positions: now Block is to there being four court verdicts, obtains the decision value of its affiliated four Tile, when having a cloud and an atural object in four decision values at least, then the decision value of Block is uncertain; When the number being judged to cloud in four decision values is four, then the decision value of Block is for very to determine cloud; When the number being judged to atural object in four decision values is four, then the decision value of Block is for very to determine atural object; When only comprise cloud in four decision values and uncertain time, then the decision value of Block is more accurate cloud; When only comprise atural object in four decision values and uncertain time, then the decision value of Block is more accurate atural object.
After obtaining the decision value of each Block, form Block index matrix, for the context correction of subsequent step.
Step S22, calculate Block gray average: the piecemeal gray average matrix utilizing the gray average composition of the corresponding each piecemeal Tile of the former figure in the first step, find the Tile piecemeal gray average comprising it that each Block is corresponding, then getting these gray averages, to obtain mean value be exactly gray average corresponding to this Block, is exactly the Block gray average matrix of lap according to the matrix of sequence of positions composition.
3rd step, expansion adjudicate the neighborhood of target accurately:
This step utilizes decision value to be the Block determining very much cloud and determine very much atural object, and revising decision value is uncertain Block decision value.
The operation of this step judges successively to revise fritter from left to right, from top to bottom, if its judgement very accurately (comprises and very determines cloud, very determine atural object), then the court verdict of the Block being judged to the more accurate type of corresponding class in each Block on its eight neighborhood is become and determine very much.Such as, a Block is judged to cloud very accurately, has with each Block on its eight neighborhood the object being judged to cloud more accurately, then the court verdict of this Block on eight neighborhood is become cloud very accurately from cloud more accurately, in like manner, for the operation that atural object is also similar.
For a judgement for very determining the Block of cloud, the court verdict of each Block on its eight neighborhood can only be determine very much cloud, more accurate cloud or uncertain these three kinds.This is because adjacent Block necessarily can comprise by same Tile, in previous step like this both them, four Tile decision values have at least should be once identical, but determines very much not to be judged in four Tile decision values of atural object and more accurate atural object cloud.In like manner, also can only have around the Block determining very much atural object and determine very much atural object, more accurate atural object or uncertain.
Wherein, for determining that very much the Block of cloud is adjacent and determining that very much the situation that the Block of cloud is adjacent appears at the inside of a continuum cloud or the inside of atural object usually, and when considering these territories, cloud sector to atural object region transfers, being spatially judged to and determining that very much the Block of cloud needs could arrive by the Block of more accurate cloud, uncertain, more accurate these types of atural object the Block being judged to and determining very much atural object.Namely more accurate cloud, uncertain, more accurate atural object are the intersections being in cloud and land, that side of intersection cloud should be that those are judged to the Block of more accurate cloud obviously, that side of atural object is then the Block being judged to more accurate atural object, if above-mentioned boundary line just in time divides Block equally, then its medium cloud is suitable with atural object content, will be judged to uncertain.
In sum, a judgement is the Block of the cloud determined very much, the object being judged to cloud is more accurately had in Block around it, these Block are just in that side of intersection cloud, owing to being subject to the impact of boundary opposite side atural object, some Tile comprising these Block also can comprise the atural object of opposite side simultaneously, Tile decision value is made to become uncertain, and then make should be determine that very much the Block of cloud becomes more accurate cloud, so its court verdict can be reduced to the cloud determined very much here; In like manner, atural object is also operated similarly.
4th step, application target context and scene context:
What utilize previous step arrives to obtain the background information that reads of result and the first step, mainly utilize the target context between adjacent Block, and reject the various false-alarm existed and the associated confidence adjusting handling object near judgement accurate target in conjunction with the scene context between Block and background information.
In this step, application target context and the contextual method of scene comprise the following step carried out in order:
Step S41, judges the background of handling object: utilize the background information obtained in the first step, the atural object thresholding that namely former figure is corresponding and Yunmen limit, analyze the size of wherein atural object thresholding, when being less than the sea route boundary thresholding of a manual setting, former figure background thinks sea, otherwise is land.Corresponding background is the situation in sea, and Yunmen limit wants suitable reduction, otherwise Yunmen limit should raise.And the rule of rising and reduction needs artificially to determine, such as: the sea route boundary thresholding deducting manually setting with the atural object thresholding in background information, then addition is limit with this result and Yunmen.Corresponding background is the situation in sea, and atural object thresholding is less than sea route boundary thresholding, so the result of subtracting each other is negative, is added just is equivalent to the thresholding reducing cloud by this result and Yunmen limit; In like manner corresponding background is the situation on land, and the thresholding of cloud has then been enhanced.
Step S42, reject the false-alarm of atural object intersection: utilize the Block decision value matrix and Block gray average matrix that obtain in second step, consider the logical relation between the eight neighborhood that Block is adjacent, if the decision value of central Block is thing very exactly, and to there is adjacent Block decision value be uncertain, more accurate cloud or more accurate atural object, and its gray average is not more than central Block's, then this adjacent Block also must be atural object, is modified to by its decision value and very determines atural object.Because the atural object component basic simlarity of adjacent isles, if itself be cloud, gray average is bound to more many than the height of central fritter, so it must be atural object.
Step S43, reject the false-alarm of cloudlet: be utilize the Block decision value matrix and Block gray average matrix that obtain in second step equally, and consider the logical relation between the eight neighborhood that Block is adjacent, if the decision value of central Block is uncertain, more accurate cloud or more accurate atural object, and eight Block major parts are around when being all judged to atural object (such as: eight the insides have be more than or equal to six and be judged to atural object), it may be just cloudlet.Because cloudlet is not clearly to blocking of atural object, so take tighter cloud gray scale thresholding to this central Block, when only having the gray scale of this central Block to be greater than this thresholding, be just judged to cloud, thus the decision value of this central Block is modified to very determines cloud.
Step S44, the adjacent target of the accurate object of adjustment judgement: the mankind are to the absolute brightness identification of object not only only being depended on to object, and relative brightness simultaneously also can have an impact to judgement.Suppose there is the uncertain target of a judgement, the higher object of it and brightness is adjacent, due to the effect of relative brightness, this target seem can be lower than it and brightness object adjacent time more secretly, certainly also just follow close to atural object.In like manner, when this target is adjacent with the lower object of brightness, seem just brighter, closer to cloud.So the uncertain target on the side of brighter cloud is easier to be considered to atural object, be in like manner easier to be considered to cloud in the uncertain target on the side of darker atural object.So, similar with previous step, in the Block of 3 × 3 sizes, if central Block is judged to very determine cloud, illustrate that this Block should be that brighter cloud, the degree of confidence that then adjacent uncertain fritter is judged to atural object should improve a grade, and namely uncertain strain is more accurate atural object.In like manner, if the decision value of central Block is for very to determine atural object, illustrate that this Block should be that darker atural object, the degree of confidence of corresponding cloud should improve a grade, and namely uncertain strain is cloud more accurately, then enters next step S45 stage;
Step S45, the process of scene context is utilized to be left target: to be judged to uncertain, to compare cloud or the atural object determined Block to remaining through step above, Yunmen limit obtained in its gray average and step S41 obtained in second step is utilized to compare, if be greater than, be judged to and very determined cloud, otherwise be judged to and very determine atural object, so far all Block are all judged to and very determine atural object, namely atural object is judged to, or be judged to and very determine cloud, be namely judged to cloud, whole process terminates.
In sum, these are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (1)
1. sentence a modified result method based on contextual remote sensing image cloud, it is characterized in that, comprising:
The first step, reads in pending related data: comprise the block index matrix of former figure, piecemeal gray average matrix and background information;
Described piecemeal is the stepping adopting the square of L × L to carry out level and vertical direction in former figure, stepping-in amount is L/2, obtain multiple piecemeal, this piecemeal is made to be Tile, every four adjacent overlapped Tile piecemeal can regard nine nonoverlapping sizes as is up and down that the little piecemeal of L/2 × L/2 is formed, and this little piecemeal is called Block;
Element in described block index matrix is the pattern-recognition decision value of each Tile piecemeal, comprises cloud, atural object and uncertain;
Element in described piecemeal gray average matrix is the gray average of each Tile piecemeal;
Described background information comprises atural object thresholding and Yunmen limit, and both is all the global threshold of corresponding former figure;
Second step, the context that application overlap partition produces, specifically comprises step S21 ~ S22;
Step S21, utilizes the decision value of Tile to produce the decision value of Block, forms Block index matrix; Divide three kinds of situations;
When Block is in the position at four angles, obtain the decision value of unique Tile belonging to it, if this decision value is cloud or atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain;
When Block is in the position on four limits: the decision value obtaining its affiliated two Tile, if two decision values are all clouds or are all atural object, then the decision value of Block is the cloud or atural object determined very much, otherwise, be judged to uncertain;
When Block is in other positions: the decision value obtaining its affiliated four Tile, when having a cloud and an atural object in four decision values at least, then the decision value of Block is uncertain; When the number being judged to cloud in four decision values is four, then the decision value of Block is for very to determine cloud; When the number being judged to atural object in four decision values is four, then the decision value of Block is for very to determine atural object; When only comprise cloud in four decision values and uncertain time, then the decision value of Block is more accurate cloud; When only comprise atural object in four decision values and uncertain time, then the decision value of Block is more accurate atural object;
Step S22, sets up Block gray average matrix according to piecemeal gray average matrix, the corresponding each Block of its element, and element value is the gray average of Block, adopts the gray average of each Tile belonging to the Block that extracts from piecemeal gray average matrix to be averaged again and obtains;
3rd step, utilizes decision value to be the Block determining very much cloud and determine very much atural object, and revising decision value is uncertain Block decision value:
The Block with eight neighborhoods is traveled through successively by order from left to right, from top to bottom, if the court verdict of each Block of the more accurate type of corresponding class on its eight neighborhood for very determining cloud or determining very much atural object, then becomes and determines very much by the decision value of Block;
4th step, application target context and scene context, specifically comprise step S41 ~ S45:
Step S41, judges the background of handling object: when the atural object thresholding in background information is less than the sea route boundary thresholding of setting, background thinks sea, otherwise is land; Corresponding background is the situation in sea, reduces Yunmen limit, otherwise, raise Yunmen limit; The modification rule of Yunmen limit is artificially determined;
Step S42, reject the false-alarm of atural object intersection: in the Block of 3 × 3 sizes, the decision value of central authorities Block is thing very definitely, and to there is adjacent Block decision value be uncertain, more accurate cloud or more accurate atural object, and its gray average is not more than central Block's, then the decision value of this adjacent Block is modified to and very determines atural object;
Step S43, reject the false-alarm of cloudlet: in the Block of 3 × 3 sizes, if the decision value of central Block is uncertain, more accurate cloud or more accurate atural object, and eight Block major parts around are all judged to atural object, and the gray average of central Block be greater than one setting cloud gray scale thresholding time, then the decision value of central Block is modified to and very determines cloud;
Step S44, the adjacent target of the accurate object of adjustment judgement: in the Block of 3 × 3 sizes, if the decision value of central Block is for very to determine cloud, then adjacent uncertain Block is corrected for more accurate atural object, in like manner, if the decision value of central Block is for very to determine atural object, then adjacent uncertain Block is corrected for more accurate cloud;
Step S45, the process of scene context is utilized to be left target: to the Block being still judged to uncertain, more accurate cloud or more accurate atural object through step above, the Yunmen obtained in the gray average of this Block and step S41 limit is compared, if be greater than, then this Block is judged to and very determines cloud, otherwise is judged to and very determines atural object; So far whole process terminates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310043710.8A CN103150582B (en) | 2013-02-04 | 2013-02-04 | Modified result method is sentenced based on contextual remote sensing image cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310043710.8A CN103150582B (en) | 2013-02-04 | 2013-02-04 | Modified result method is sentenced based on contextual remote sensing image cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103150582A CN103150582A (en) | 2013-06-12 |
| CN103150582B true CN103150582B (en) | 2015-11-25 |
Family
ID=48548644
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310043710.8A Active CN103150582B (en) | 2013-02-04 | 2013-02-04 | Modified result method is sentenced based on contextual remote sensing image cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103150582B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490225B (en) * | 2019-07-09 | 2022-06-28 | 北京迈格威科技有限公司 | Scene-based image classification method, device, system and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101221662A (en) * | 2008-01-31 | 2008-07-16 | 复旦大学 | Remote sensing image mixed image element decomposition method based on self-organizing mapping neural network |
| CN101246545A (en) * | 2008-02-22 | 2008-08-20 | 华南师范大学 | Possion method for removing cloud from optical remote sensing image |
| CN102663348A (en) * | 2012-03-21 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Marine ship detection method in optical remote sensing image |
| CN102750701A (en) * | 2012-06-15 | 2012-10-24 | 西安电子科技大学 | Method for detecting spissatus and spissatus shadow based on Landsat thematic mapper (TM) images and Landsat enhanced thematic mapper (ETM) images |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8189877B2 (en) * | 2005-10-21 | 2012-05-29 | Carnegie Institution Of Washington | Remote sensing analysis of forest disturbances |
-
2013
- 2013-02-04 CN CN201310043710.8A patent/CN103150582B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101221662A (en) * | 2008-01-31 | 2008-07-16 | 复旦大学 | Remote sensing image mixed image element decomposition method based on self-organizing mapping neural network |
| CN101246545A (en) * | 2008-02-22 | 2008-08-20 | 华南师范大学 | Possion method for removing cloud from optical remote sensing image |
| CN102663348A (en) * | 2012-03-21 | 2012-09-12 | 中国人民解放军国防科学技术大学 | Marine ship detection method in optical remote sensing image |
| CN102750701A (en) * | 2012-06-15 | 2012-10-24 | 西安电子科技大学 | Method for detecting spissatus and spissatus shadow based on Landsat thematic mapper (TM) images and Landsat enhanced thematic mapper (ETM) images |
Non-Patent Citations (1)
| Title |
|---|
| 遥感图像的实时云判技术;闫宇松;《北京理工大学学报》;20100731;第30卷(第7期);817-821 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103150582A (en) | 2013-06-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6966989B2 (en) | Methods, devices, electronics, storage media and computer programs for determining lane boundaries on the road | |
| Wei et al. | Toward automatic building footprint delineation from aerial images using CNN and regularization | |
| WO2017215622A1 (en) | Object segmentation method and apparatus and computing device | |
| CN111062413A (en) | Road target detection method and device, electronic equipment and storage medium | |
| CN110378297B (en) | Remote sensing image target detection method and device based on deep learning and storage medium | |
| CN109447994A (en) | In conjunction with the remote sensing image segmentation method of complete residual error and Fusion Features | |
| CN106228125B (en) | Method for detecting lane lines based on integrated study cascade classifier | |
| Wu et al. | Automatic road extraction from high-resolution remote sensing images using a method based on densely connected spatial feature-enhanced pyramid | |
| CN110659664B (en) | SSD-based high-precision small object identification method | |
| Medina-Carnicer et al. | Unimodal thresholding for edge detection | |
| CN103106667A (en) | Motion target tracing method towards shielding and scene change | |
| CN103473763B (en) | Road edge detection method based on heuristic Probabilistic Hough Transform | |
| CN110543906B (en) | Automatic skin recognition method based on Mask R-CNN model | |
| CN114897779A (en) | Cervical cytology image abnormal area positioning method and device based on fusion attention | |
| CN105160355A (en) | Remote sensing image change detection method based on region correlation and visual words | |
| CN111915657A (en) | Point cloud registration method and device, electronic equipment and storage medium | |
| CN113159215A (en) | Small target detection and identification method based on fast Rcnn | |
| You et al. | Lane detection algorithm for night-time digital image based on distribution feature of boundary pixels | |
| Liu et al. | A new multi-channel deep convolutional neural network for semantic segmentation of remote sensing image | |
| CN105046278A (en) | Optimization method of Adaboost detection algorithm on basis of Haar features | |
| CN116051822A (en) | Concave obstacle recognition method and device, processor and electronic equipment | |
| Zhao et al. | Monocular pedestrian orientation estimation based on deep 2D-3D feedforward | |
| CN106600613A (en) | Embedded GPU-based improved LBP infrared target detection method | |
| CN103150582B (en) | Modified result method is sentenced based on contextual remote sensing image cloud | |
| CN110874170A (en) | Image area correction method, image segmentation method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |