CN102867196A - Method for detecting complex sea-surface remote sensing image ships based on Gist characteristic study - Google Patents
Method for detecting complex sea-surface remote sensing image ships based on Gist characteristic study Download PDFInfo
- Publication number
- CN102867196A CN102867196A CN2012103397911A CN201210339791A CN102867196A CN 102867196 A CN102867196 A CN 102867196A CN 2012103397911 A CN2012103397911 A CN 2012103397911A CN 201210339791 A CN201210339791 A CN 201210339791A CN 102867196 A CN102867196 A CN 102867196A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- image
- feature
- brightness
- color
- 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.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting complex sea-surface remote sensing image ships based on Gist characteristic study, comprising the following steps: step 1, collecting the remote sensing image data of complex sea-surfaces which have different time phases, different sensors and different sizes; step 2, implementing the block preprocess for the complex sea-surface remote sensing image to obtain a sample image section and a detection image section; step 3, drawing the distinguishing features and the Gist features of the sample image section and the detection image section; step 4, training the sample image section according to the distinguishing features and the Gist features obtained in the step 3 to obtain a training mode; step 5, adopting an SVM (Support Vector Machine) classifier to judge whether the detection image section has ships according to the training mode obtained in the step 4; and step 6, finding the single ship of the detection image section based on an improved itti visual attention mode. The method for detecting the complex sea-surface remote sensing image ships based on Gist characteristic study ensures that the ships do not fail to examine, reduces the false alarm rate, effectively treats the complex sea-surface remote sensing images under the disturbed conditions of sea clutters, cloud and mist, and has low computation complexity and strong pertinence.
Description
Technical field
The present invention relates to the Remote Sensing Image Processing Technology field, especially relate to a kind of complicated sea remote sensing image Ship Detection based on Gist biological vision feature learning.
Background technology
Ship Target Detection is the task with traditional of each coastal strip country of the world, has a wide range of applications in naval vessel searching and relief, fisherman supervision, illegal immigrant, defendance territory, the aspects such as Monitoring and Management anti-drug, naval vessel illegal dumping greasy dirt.Along with the development of remotely sensed image technology, utilizing remote sensing images to carry out Ship Target Detection becomes possibility, and its research object comprises the detection to naval vessel itself and ship Wake.Under the environmental baseline of complicated ocean, what present rambling fish scale light, large tracts of land retroreflective regions, irregular movement on the satellite remote-sensing image enriches texture wave etc., middle-size and small-size Ship Target may be hidden in the complicated background clutter, thereby affects the Ship Target recognition effect.Sea situation on the Ship Target image (stormy waves), meteorological (cloud and mist), water colour etc. cause the interference such as remote sensing image Zhong Hai land characteristic is unstable, wave of the sea, floating on water thing, ship Wake very many, and the naval vessel detection difficulty is large.
Existing Ship Target Detection mainly contains following algorithm: the algorithm of cutting apart based on global threshold, the algorithm based on Local threshold segmentation, best entropy automatic threshold method, based on the algorithm of distributed model, based on the algorithm of fractal model, based on algorithm of target detection of property field etc., but their ubiquities detect the too high problem of false alarm rate, and universality is relatively poor.Wherein, the global threshold algorithm can't be regulated threshold value automatically according to the variation of regional area in the image, therefore testing result is subject to localized variation and introduces a large amount of false-alarms and undetected, the image wider to fabric width, reason owing to imaging mechanism, Sea background has strong grey scale change, also can cause occurring false-alarm and undetected.These class methods have only been utilized the gray-scale statistical characteristic of target in testing process, do not consider the space structure information of target, and contacting of histogram shape and picture material also has uncertainty; The local threshold algorithm spot make an uproar more, when the sea stormy waves is larger, cause easily a large amount of false-alarms.This algorithm is owing to need repeatedly the statistical parameter of background area in the calculation window simultaneously, and operand is very big, and the speed of processing procedure is slower, can not satisfy the real-time or near needs of processing in real time in the practical application; Choosing of best entropy automatic threshold method threshold value only utilized the gray-scale statistical characteristic, do not consider the space structure information of target, and contacting of histogrammic distribution and picture material also has uncertainty.In traditional KSW algorithm, criterion function is defined as target gray scale entropy and background gray scale entropy sum simply, and background gray scale entropy and target gray scale entropy occupy equal ratio in criterion function.This has ignored background and target proportion different on image, has also ignored background and the target difference on tonal range.When image comprises stronger extra large clutter, the testing result that often can not get; Algorithm based on distributed model at first needs background is done the clutter distributional assumption, and this just needs certain priori, but background clutter is not strictly obeyed certain distribution yet in fact generally speaking.Secondly, such algorithm need to be added up each pixel in the image, so calculated amount is larger, and increases along with the increase of sliding window size; Think that based on the algorithm of fractal model the fractal dimension of natural scene and Ship Target has certain difference, detects according to difference.But affected by background complexity, random noise, image quality etc., single yardstick or constant fractal dimension are difficult to distinguish natural scene and man-made target; Based on the algorithm of target detection of property field, when the intensity profile of background is complicated, affected by noise greatly can increase noise to the impact of characteristic pattern this moment, produces mistake and cut apart.In addition, also some impacts can be arranged on the profile of target itself in the process of Feature Conversion, utilize the method segmentation object, shape information will be lost to some extent.
Can find out that various algorithms still will be subject to the restriction of many conditions, be subjected to the impact of weather, illumination variation such as interference, the target of image background, particularly for the remotely-sensed data of middle low resolution, Ship Target is rendered as little target at image, and the probability that occurs false dismissal, false-alarm in the observation process is higher.Simultaneously, in the ordinary course of things, visible images can be subject to the interference such as cloud layer and greasy dirt, wave, set up relatively difficulty of background model, target and the background subtraction opposite sex are inconsistent in the image, and the naval vessel intensity profile is inhomogeneous in the image, also not obvious with extra large background contrast, therefore also be difficult for cutting apart, for black polarity naval vessel, all the more so especially.
Summary of the invention
For the problems referred to above, the present invention proposes a kind of complicated sea remote sensing image Ship Detection based on the Gist feature learning.
Technical scheme of the present invention is a kind of complicated sea remote sensing image Ship Detection based on the Gist feature learning, may further comprise the steps:
Step 1, the data of collection complicated sea remote sensing image;
Step 2 is carried out partitioning pretreatment to the complicated sea remote sensing image, obtains the sample image section and detects image slice;
Step 3 is extracted notable feature and Gist feature that sample image is cut into slices and detected image slice;
Step 4, according to step 3 gained notable feature and Gist feature to sample image section train, obtain training pattern;
Step 5 according to step 4 gained training pattern, judges with the svm classifier device whether arbitrary detection image slice contains the naval vessel, is then to enter step 5, otherwise finishes this is detected the detection of image slice;
Step 6 is sought the single naval vessel that detects image slice based on the itti visual attention model, obtains ship target.
And, in the step 3 the notable feature extraction is carried out in arbitrary sample image section and detection image slice, implementation is as follows,
Image slice is decomposed on color, brightness, these 3 feature passages of direction, and color, brightness passage represent with gaussian pyramid respectively, by the characteristic pattern on the poor operation generation of the central peripheral color characteristic passage and the characteristic pattern on the brightness passage; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, the direction pyramid of 135 ° of 4 directions generates characteristic pattern on each direction by the poor operation of central peripheral;
Merge by local nonlinearity and each feature Graph Character subgraph to be fused to color is significantly schemed, brightness is significantly schemed and direction is significantly schemed, represent at last the notable feature of image slice with distance between average, standard deviation, Local modulus maxima and the Local modulus maxima of each remarkable figure.
And, in the step 2 described Gist feature extraction is carried out in arbitrary sample image section and detection image slice, implementation is as follows,
Each image slice is decomposed on color, brightness, 3 feature passages of direction, color, brightness passage are used respectively this pyramid representation of 9 floor heights, by the characteristic pattern on the poor operation generation of the central peripheral color characteristic passage and the characteristic pattern on the brightness passage; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, the direction pyramid of 135 ° of 4 directions obtains the characteristic pattern on each direction passage;
The sub-grid that every feature subgraph of each characteristic pattern is divided into 4 * 4 sizes, average after asking respectively 4 sub-grids in average, the upper left corner, 4 sub-grids in the upper right corner, 4 sub-grids in the lower left corner, 4 sub-grids in the upper left corner, 4 sub-grids in the upper left corner of 16 sub-grids to merge respectively and the average of whole feature subgraph represent the Gist proper vector with acquired results at last.
And, seek the single naval vessel that detects image slice based on improved itti visual attention model in the step 6, may further comprise the steps,
Step 6.1 is carried out the low-level visual features extraction to detecting image slice, comprises brightness, color, direction and textural characteristics;
Step 6.2, significantly figure generates, comprise that 4 features of brightness, color, direction, texture that step 6.1 is obtained generate each feature subgraph by the poor operation of central peripheral, then merge to generate with the local nonlinearity fusion method that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture is significantly schemed, merge with the local nonlinearity fusion method at last to generate whole significantly figure; Merge to generate that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture significantly before the figure each pixel value to the individual features subgraph carry out square operation;
Step 6.3, Ship Target Detection comprises that the remarkable figure to obtaining in the step 6.2 passes through the single naval vessel that the inhibition of return technology for detection detects image slice, obtains ship target.
And, select to derive from not the simultaneously remote sensing image data of phase, different sensors, different scale when gathering the data of complicated sea remote sensing image, after the complicated sea remote sensing image carried out partitioning pretreatment, the negative sample image that mark contains the positive sample image on naval vessel and do not contain the naval vessel was cut into slices as sample image.
The method that the present invention proposes is at first carried out piecemeal to image, extract notable feature and the Gist feature of each sub-image piece, according to the proper vector that extracts sample data is trained, then utilize the svm classifier device to judge and whether contain the naval vessel in the sub-image piece, at last based on the single naval vessel in the improved itti visual attention model searching sub-image piece.The present invention is its false alarm rate of reduce when having guaranteed not undetected naval vessel, can effectively process to contain the complicated sea remote sensing image that extra large clutter, cloud and mist etc. disturb, and computation complexity is low, with strong points; Be applicable to not the simultaneously complicated sea remote sensing image data of phase, different resolution, have good universality; Can carry out to the magnanimity remote sensing image data of large format fast processing and detect the naval vessel.
Description of drawings
Fig. 1 is process flow diagram of the present invention.
Embodiment
Describe technical solution of the present invention in detail below in conjunction with drawings and Examples.
The complicated sea remote sensing image Ship Detection that the present invention is based on Gist biological vision feature learning is notable feature and the Gist feature of utilizing the naval vessel, sub-image piece to jumbo image is trained with svm classifier device (support vector machine classifier), obtain a forecast model, support vector in this model has represented the characteristic feature on naval vessel, according to improved itti visual attention model the sub-image piece that has doubtful naval vessel to exist is carried out the detection on single naval vessel again.The Gist feature can referring to existing document " Driving me Around the Bend:Learning to Drive from Visual Gist ", can be utilized biological vision feature calculation Gist feature.The embodiment flow process can adopt computer software technology to realize automatically operation, as shown in Figure 1, specifically may further comprise the steps:
Step 1, the data of collection complicated sea remote sensing image.
During implementation, adopting training data to cover should be many as much as possible, with the detecting step of reply back.The embodiment collection is the complicated sea remote sensing image data of phase, different sensors, different scale simultaneously not.
Step 2 is carried out partitioning pretreatment to the complicated sea remote sensing image, makes sample data, obtains the sample image section and detects image slice.
Embodiment processes remote sensing image block, is to be equal and opposite in direction with original jumbo image block, the image slice of N * N size, i.e. and each section comprises N * N pixel.
With gather not simultaneously behind the complicated sea remote sensing image data piecemeal of phase, different sensors, different scale, from the gained image slice, can select training sample, open positive sample image section (naval vessel) and S opens negative sample image slice (non-naval vessel) by handmarking S, jointly consist of the training set that sample image is cut into slices.Perhaps can directly adopt in advance the sample image section of mark.Other image slice can be as detection image slice to be sorted.Generally all be with view picture complicated sea remote sensing image as image to be detected, behind image block to be detected, all sub-image pieces of gained are as detection image slice to be sorted.
Wherein, the concrete value of N and S can according to circumstances be set voluntarily by those skilled in the art.
Step 3 is extracted notable feature and Gist feature that sample image is cut into slices and detected image slice.
Among the embodiment, the notable feature extraction is carried out in arbitrary sample image section and detection image slice, specifically:
Image slice is decomposed on color, brightness, 3 feature passages of direction, color, brightness passage are used respectively this pyramid representation of 9 floor heights, generate respectively characteristic pattern on the color characteristic passage and the characteristic pattern on the brightness passage by the poor operation of central peripheral; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, 9 layers of direction pyramid of 135 ° of 4 directions generate characteristic pattern on each direction by the poor operation of central peripheral respectively.Wherein, characteristic pattern on the color characteristic passage comprises that 12 color characteristic subgraphs (divide two groups, one group is red green color contrast, one group is the champac color contrast, every group each 6), characteristic pattern on the brightness passage comprises 6 brightness subgraphs (becoming 6 after the poor operation of 9 layers of pyramids process central peripheral), characteristic pattern on each direction of direction character passage comprises 6 direction character subgraphs (9 layers of pyramid through the poor operation of central peripheral then become 6), 24 feature subgraphs altogether on 4 directions.Altogether 42 feature subgraphs, i.e. 42=12+6+4 * 6.Then merge by local nonlinearity and characteristic pattern is fused to color, brightness, direction significantly schemes, wherein 12 color characteristic subgraphs are fused to 1 color and significantly scheme, 6 brightness subgraphs are fused to 1 brightness significantly schemes, totally 24 direction character subgraphs on 4 directions are fused to 1 direction significantly schemes, totally 3 remarkable figure.At last with image slice each significantly figure minute other average, standard deviation, Local modulus maxima, local maximum dot spacing from, with 3 significantly 12 dimensional vectors of figure minute other 4 values formations represent the notable feature of image slice.The average of image, standard deviation, Local modulus maxima, the local maximum dot spacing is from specifically asking for, the poor operation of central peripheral and local non-linear fusion are prior art, can be referring to document " L.Itti; C.Koch; E.Niebur; AModel of Saliency-Based Visual Attention for Rapid Scene Analysis; IEEE Transactions on Pattern Analysis and Machine Intelligence; Vol.20, No.11, pp.1254-1259, Nov 1998. "; " L.Itti, C.Koch, Comparison of Feature Combination Strategies for Saliency-Based Visual Attention Systems, In:Proc.SPIE Human Vision and Electronic Imaging IV (HVEI'99), San Jose, CA, Vol.3644, pp.473-82, Bellingham, WA:SPIE Press, Jan 1999. "
Among the embodiment, the Gist feature extraction is carried out in arbitrary sample image section and detection image slice, specifically:
Image slice is decomposed on color, brightness, 3 feature passages of direction, color, brightness passage are used respectively this pyramid representation of 9 floor heights, generate respectively characteristic pattern on the color characteristic passage and the characteristic pattern on the brightness passage by the poor operation of central peripheral; The direction character passage obtains 0 ° with Gabor filtering, 45 °, 90 °, 4 layers of direction pyramid of 135 ° of 4 directions, the image (being that original image is respectively divided by 20,21,22,23 4 tomographic images that obtain) of getting 4 yardsticks of each direction pyramid as the direction character subgraph of respective direction, obtains 4 characteristic patterns on the direction.Characteristic pattern on the color characteristic passage comprises 12 color characteristic subgraphs, and the characteristic pattern on the brightness passage comprises 6 brightness subgraphs, and the characteristic pattern on each direction of direction character passage comprises 4 direction character subgraphs.Altogether 34 feature subgraphs, i.e. 34=6+12+4 * 4.Every feature subgraph is divided into 4 * 4 sub-grids, ask respectively the average of 16 sub-grids, the average of the average after 4 sub-grids in the upper left corner, 4 sub-grids in the upper right corner, 4 sub-grids in the lower left corner, 4 sub-grids in the upper left corner, 4 sub-grids in the upper left corner merge respectively (being exactly as its average of overall calculation with 4 sub-grids) and whole feature subgraph, obtain the vector of one 21 dimension, obtain at last the Gist proper vector of 34 * 21=714 dimension.
During implementation, the gaussian pyramid number of plies can according to circumstances be set voluntarily by those skilled in the art, and the number of plies is more, and the feature dimensions of then extracting is also just higher.
Step 4, according to step 3 gained notable feature and Gist feature to sample image section train, obtain training pattern.
According to the notable feature of sample image section, and belong to positive sample image section or negative sample image slice, can train.Concrete training realizes adopting SVM training aids of the prior art, can obtain training pattern after all sample image sections are trained.
Step 5 according to step 4 gained training pattern, judges to detect in the image slice whether contain the naval vessel with the svm classifier device.For arbitrary detection image slice, if judged result enters step 6 for containing, if judged result finishes this is detected the processing of image slice for not containing.
Training pattern realizes by existing SVM method all that with predicting the outcome for example the SVM kernel function is with having RBF kernel function (radial basis function) now, and it will not go into details in the present invention.
Step 6 is sought the single naval vessel that detects image slice based on the itti visual attention model, obtains ship target.
Itti visual attention model specific implementation is prior art, and for the purpose of improving Detection accuracy, embodiment further proposes the itti visual attention model is improved, and seeks the single naval vessel that detects image slice based on improved itti visual attention model.Described improved itti visual attention model is to have increased textural characteristics on the basis of original itti model, expanded its characteristic range, before the feature subgraph of different scale merges, carry out a square of stretching computing simultaneously, strengthen the difference of significantly distinguishing with non-remarkable district, outstanding Ship Target.
The step of embodiment is as follows:
Step 6.1 is carried out the low-level visual features extraction to detecting image slice, comprises brightness, color, direction and textural characteristics.Original itti visual attention model extracts respectively brightness, color, direction as low-level visual features, embodiment has increased texture feature extraction, can realize texture feature extraction by existing discrete square converter technique, referring to document " VD.Gesu; C.Valent; L.Strinati.Local operators to detect regions of interest[J] .Pattern Recognition Letter, 1997,18 (11-13): 1077-1081. "
The formula of the discrete square conversion (DMT) of prior art
In the formula, i, j detect ranks value in the image slice, and g (i-r, j-s) expression detects the original pixel value of the capable j-s row of i-r in the image slice; K is the window size parameter of presetting, k=1 for example, and window is 3 * 3; R, s are the loop variables in the window, and p, q are indexes, and those skilled in the art can set k, r, s value voluntarily as required during implementation.Embodiment gets respectively (p=0, q=1) (p=1, q=0), and (p=1, q=1) i.e. be three DMT textural characteristics, DMT
1.0DMT
0.1DMT
1.1
Step 6.2, significantly figure generates, comprise that 4 features of brightness, color, direction, texture that step 6.1 is obtained generate each feature subgraph by the poor operation of central peripheral, then merge to generate with the local nonlinearity fusion method that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture is significantly schemed, merge with the local nonlinearity fusion method at last and generate whole significantly figure, merge generate that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture significantly before the figure each pixel value to the individual features subgraph carry out square operation.Because square operation, the scope of pixel value among the remarkable figure that stretched so that the marking area layering is more obvious, has been given prominence to Ship Target.
Among the embodiment, color, brightness passage are used respectively this pyramid representation of 9 floor heights, generate respectively characteristic pattern on the color characteristic passage and the characteristic pattern on the brightness passage by the poor operation of central peripheral; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, 9 layers of direction pyramid of 135 ° of 4 directions generate characteristic pattern on each direction by the poor operation of central peripheral.Wherein, characteristic pattern on the color characteristic passage comprises 12 color characteristic subgraphs, characteristic pattern on the brightness passage comprises 6 brightness subgraphs, and the characteristic pattern on each direction of direction character passage comprises 6 direction character subgraphs, 24 feature subgraphs altogether on 4 directions.Altogether 42 feature subgraphs, i.e. 42=12+6+4 * 6.Then with the feature subgraph of all different scales as input, merge by local nonlinearity and characteristic pattern to be fused to color, brightness, direction significantly to scheme, wherein 12 color characteristic subgraphs are fused to 1 color and significantly scheme, 6 brightness subgraphs are fused to 1 brightness significantly schemes, totally 24 direction character subgraphs on 4 directions are fused into 1 direction significantly schemes, totally 3 remarkable figure.
Characteristic pattern on the textural characteristics passage that the discrete square conversion of employing obtains comprises 3 textural characteristics subgraphs, 3 textural characteristics subgraphs are fused to 1 texture significantly scheme, 3 characteristic remarkable pictures before adding are significantly schemed for 4 altogether, they are fused to 1 integral body significantly scheme, comprehensively all features.
Step 6.3, Ship Target Detection, the remarkable figure that obtains in step 6.2 detect the single Ship Target of image slice by the inhibition of return technology for detection.It is existing that the inhibition of return technology can it will not go into details in the present invention referring to " Nov 1998. for L.Itti; C.Koch; E.Niebur; A Model of Saliency-Based Visual Attention for Rapid Scene Analysis; IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.20, No.11; pp.1254-1259 ".
Specific embodiment described herein only is to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or replenish or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.
Claims (5)
1. the complicated sea remote sensing image Ship Detection based on the Gist feature learning is characterized in that, may further comprise the steps:
Step 1, the data of collection complicated sea remote sensing image;
Step 2 is carried out partitioning pretreatment to the complicated sea remote sensing image, obtains the sample image section and detects image slice;
Step 3 is extracted notable feature and Gist feature that sample image is cut into slices and detected image slice;
Step 4, according to step 3 gained notable feature and Gist feature to sample image section train, obtain training pattern;
Step 5 according to step 4 gained training pattern, judges with the svm classifier device whether arbitrary detection image slice contains the naval vessel, is then to enter step 5, otherwise finishes this is detected the detection of image slice;
Step 6 is sought the single naval vessel that detects image slice based on the itti visual attention model, obtains ship target.
2. described complicated sea remote sensing image Ship Detection based on the Gist feature learning according to claim 1 is characterized in that: to arbitrary sample image section with detect image slice and carry out notable feature and extract, implementation is as follows in the step 3,
Image slice is decomposed on color, brightness, these 3 feature passages of direction, and color, brightness passage represent with gaussian pyramid respectively, by the characteristic pattern on the poor operation generation of the central peripheral color characteristic passage and the characteristic pattern on the brightness passage; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, the direction pyramid of 135 ° of 4 directions generates characteristic pattern on each direction by the poor operation of central peripheral;
Merge by local nonlinearity and each feature Graph Character subgraph to be fused to color is significantly schemed, brightness is significantly schemed and direction is significantly schemed, represent at last the notable feature of image slice with distance between average, standard deviation, Local modulus maxima and the Local modulus maxima of each remarkable figure.
3. described complicated sea remote sensing image Ship Detection based on the Gist feature learning according to claim 1 is characterized in that: to arbitrary sample image section with detect image slice and carry out described Gist feature extraction, implementation is as follows in the step 2,
Each image slice is decomposed on color, brightness, 3 feature passages of direction, color, brightness passage are used respectively this pyramid representation of 9 floor heights, by the characteristic pattern on the poor operation generation of the central peripheral color characteristic passage and the characteristic pattern on the brightness passage; The direction character passage obtains 0 ° with Gabor filtering, and 45 °, 90 °, the direction pyramid of 135 ° of 4 directions obtains the characteristic pattern on each direction passage;
The sub-grid that every feature subgraph of each characteristic pattern is divided into 4 * 4 sizes, average after asking respectively 4 sub-grids in average, the upper left corner, 4 sub-grids in the upper right corner, 4 sub-grids in the lower left corner, 4 sub-grids in the upper left corner, 4 sub-grids in the upper left corner of 16 sub-grids to merge respectively and the average of whole feature subgraph represent the Gist proper vector with acquired results at last.
4. described complicated sea remote sensing image Ship Detection based on the Gist feature learning according to claim 1 is characterized in that: seeks the single naval vessel that detects image slice based on improved itti visual attention model in the step 6, may further comprise the steps,
Step 6.1 is carried out the low-level visual features extraction to detecting image slice, comprises brightness, color, direction and textural characteristics;
Step 6.2, significantly figure generates, comprise that 4 features of brightness, color, direction, texture that step 6.1 is obtained generate each feature subgraph by the poor operation of central peripheral, then merge to generate with the local nonlinearity fusion method that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture is significantly schemed, merge with the local nonlinearity fusion method at last to generate whole significantly figure; Merge to generate that brightness is significantly schemed, color is significantly schemed, direction is significantly schemed, texture significantly before the figure each pixel value to the individual features subgraph carry out square operation;
Step 6.3, Ship Target Detection comprises that the remarkable figure to obtaining in the step 6.2 passes through the single naval vessel that the inhibition of return technology for detection detects image slice, obtains ship target.
5. according to claim 1 and 2 or 3 or 4 described complicated sea remote sensing image Ship Detections based on the Gist feature learning, it is characterized in that: select to derive from not the simultaneously remote sensing image data of phase, different sensors, different scale when gathering the data of complicated sea remote sensing image, after the complicated sea remote sensing image carried out partitioning pretreatment, the negative sample image that mark contains the positive sample image on naval vessel and do not contain the naval vessel was cut into slices as sample image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210339791.1A CN102867196B (en) | 2012-09-13 | 2012-09-13 | Based on the complicated sea remote sensing image Ship Detection of Gist feature learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210339791.1A CN102867196B (en) | 2012-09-13 | 2012-09-13 | Based on the complicated sea remote sensing image Ship Detection of Gist feature learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102867196A true CN102867196A (en) | 2013-01-09 |
| CN102867196B CN102867196B (en) | 2015-10-21 |
Family
ID=47446060
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210339791.1A Active CN102867196B (en) | 2012-09-13 | 2012-09-13 | Based on the complicated sea remote sensing image Ship Detection of Gist feature learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102867196B (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177608A (en) * | 2013-03-01 | 2013-06-26 | 上海海事大学 | Offshore suspicious ship and ship oil contamination discovering system |
| CN103530628A (en) * | 2013-10-29 | 2014-01-22 | 上海市城市建设设计研究总院 | High-resolution remote sensing image ortho-rectification method based on floating control point |
| CN105243154A (en) * | 2015-10-27 | 2016-01-13 | 武汉大学 | Remote sensing image retrieval method and system based on significant point characteristics and spare self-encodings |
| CN105335765A (en) * | 2015-10-20 | 2016-02-17 | 北京航天自动控制研究所 | Method for detecting characteristic region matched with SAR |
| CN106651937A (en) * | 2016-10-19 | 2017-05-10 | 成都电科智达科技有限公司 | Method for detecting small unmanned aerial vehicle target based on super-pixels and scene prediction |
| CN106778495A (en) * | 2016-11-21 | 2017-05-31 | 北京航天宏图信息技术股份有限公司 | Ship Detection in remote sensing image under complicated sea background |
| CN107563303A (en) * | 2017-08-09 | 2018-01-09 | 中国科学院大学 | A kind of robustness Ship Target Detection method based on deep learning |
| CN107704865A (en) * | 2017-05-09 | 2018-02-16 | 北京航空航天大学 | Fleet Targets Detection based on the extraction of structure forest edge candidate region |
| CN107977621A (en) * | 2017-11-29 | 2018-05-01 | 淮海工学院 | Shipwreck identification model construction method, device, electronic equipment and storage medium |
| CN108052585A (en) * | 2017-12-11 | 2018-05-18 | 江苏丰华谷科技发展有限公司 | The determination method of dynamic object in a kind of complex environment |
| CN108133184A (en) * | 2017-12-20 | 2018-06-08 | 中国水产科学研究院渔业机械仪器研究所 | It is a kind of based on the fish of fractal theory and BP algorithm identification and vaccine injection method |
| CN109427055A (en) * | 2017-09-04 | 2019-03-05 | 长春长光精密仪器集团有限公司 | The remote sensing images surface vessel detection method of view-based access control model attention mechanism and comentropy |
| CN109558771A (en) * | 2017-09-26 | 2019-04-02 | 中电科海洋信息技术研究院有限公司 | Behavior state recognition methods, device, equipment and the storage medium of marine ships |
| CN109800716A (en) * | 2019-01-22 | 2019-05-24 | 华中科技大学 | One kind being based on the pyramidal Oceanic remote sensing image ship detecting method of feature |
| CN109871823A (en) * | 2019-03-11 | 2019-06-11 | 中国电子科技集团公司第五十四研究所 | A kind of satellite image Ship Detection of combination rotating frame and contextual information |
| CN110008833A (en) * | 2019-02-27 | 2019-07-12 | 中国科学院半导体研究所 | Target ship detection method based on remote sensing image |
| CN110084181A (en) * | 2019-04-24 | 2019-08-02 | 哈尔滨工业大学 | A kind of remote sensing images Ship Target Detection method based on sparse MobileNetV2 network |
| CN111738236A (en) * | 2020-08-14 | 2020-10-02 | 之江实验室 | Image segmentation identification method, device and system of self-adaptive hierarchy |
| CN112766083A (en) * | 2020-12-30 | 2021-05-07 | 中南民族大学 | Remote sensing scene classification method and system based on multi-scale feature fusion |
| CN113256720A (en) * | 2021-06-03 | 2021-08-13 | 浙江大学 | Method for simultaneously detecting SAR image ship and trail thereof |
| CN116310734A (en) * | 2023-04-25 | 2023-06-23 | 慧铁科技有限公司 | Fault detection method and system for railway wagon running part based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100092036A1 (en) * | 2008-06-17 | 2010-04-15 | Subhodev Das | Method and apparatus for detecting targets through temporal scene changes |
| CN102214298A (en) * | 2011-06-20 | 2011-10-12 | 复旦大学 | Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism |
| 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 |
-
2012
- 2012-09-13 CN CN201210339791.1A patent/CN102867196B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100092036A1 (en) * | 2008-06-17 | 2010-04-15 | Subhodev Das | Method and apparatus for detecting targets through temporal scene changes |
| CN102214298A (en) * | 2011-06-20 | 2011-10-12 | 复旦大学 | Method for detecting and identifying airport target by using remote sensing image based on selective visual attention mechanism |
| 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 |
Cited By (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177608A (en) * | 2013-03-01 | 2013-06-26 | 上海海事大学 | Offshore suspicious ship and ship oil contamination discovering system |
| CN103530628B (en) * | 2013-10-29 | 2016-08-17 | 上海市城市建设设计研究总院 | High-resolution remote sensing image ortho-rectification method based on floating control point |
| CN103530628A (en) * | 2013-10-29 | 2014-01-22 | 上海市城市建设设计研究总院 | High-resolution remote sensing image ortho-rectification method based on floating control point |
| CN105335765A (en) * | 2015-10-20 | 2016-02-17 | 北京航天自动控制研究所 | Method for detecting characteristic region matched with SAR |
| CN105243154B (en) * | 2015-10-27 | 2018-08-21 | 武汉大学 | Remote sensing image retrieval method based on notable point feature and sparse own coding and system |
| CN105243154A (en) * | 2015-10-27 | 2016-01-13 | 武汉大学 | Remote sensing image retrieval method and system based on significant point characteristics and spare self-encodings |
| CN106651937A (en) * | 2016-10-19 | 2017-05-10 | 成都电科智达科技有限公司 | Method for detecting small unmanned aerial vehicle target based on super-pixels and scene prediction |
| CN106651937B (en) * | 2016-10-19 | 2019-10-18 | 成都电科智达科技有限公司 | A kind of small drone object detection method based on super-pixel and scene prediction |
| CN106778495A (en) * | 2016-11-21 | 2017-05-31 | 北京航天宏图信息技术股份有限公司 | Ship Detection in remote sensing image under complicated sea background |
| CN107704865A (en) * | 2017-05-09 | 2018-02-16 | 北京航空航天大学 | Fleet Targets Detection based on the extraction of structure forest edge candidate region |
| CN107563303A (en) * | 2017-08-09 | 2018-01-09 | 中国科学院大学 | A kind of robustness Ship Target Detection method based on deep learning |
| CN107563303B (en) * | 2017-08-09 | 2020-06-09 | 中国科学院大学 | Robust ship target detection method based on deep learning |
| CN109427055A (en) * | 2017-09-04 | 2019-03-05 | 长春长光精密仪器集团有限公司 | The remote sensing images surface vessel detection method of view-based access control model attention mechanism and comentropy |
| CN109427055B (en) * | 2017-09-04 | 2022-12-20 | 长春长光精密仪器集团有限公司 | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy |
| CN109558771A (en) * | 2017-09-26 | 2019-04-02 | 中电科海洋信息技术研究院有限公司 | Behavior state recognition methods, device, equipment and the storage medium of marine ships |
| CN107977621A (en) * | 2017-11-29 | 2018-05-01 | 淮海工学院 | Shipwreck identification model construction method, device, electronic equipment and storage medium |
| CN108052585A (en) * | 2017-12-11 | 2018-05-18 | 江苏丰华谷科技发展有限公司 | The determination method of dynamic object in a kind of complex environment |
| CN108052585B (en) * | 2017-12-11 | 2021-11-23 | 江苏丰华联合科技有限公司 | Method for judging dynamic target in complex environment |
| CN108133184A (en) * | 2017-12-20 | 2018-06-08 | 中国水产科学研究院渔业机械仪器研究所 | It is a kind of based on the fish of fractal theory and BP algorithm identification and vaccine injection method |
| CN108133184B (en) * | 2017-12-20 | 2020-04-28 | 中国水产科学研究院渔业机械仪器研究所 | Fish identification and vaccine injection method based on fractal theory and BP algorithm |
| CN109800716A (en) * | 2019-01-22 | 2019-05-24 | 华中科技大学 | One kind being based on the pyramidal Oceanic remote sensing image ship detecting method of feature |
| CN110008833A (en) * | 2019-02-27 | 2019-07-12 | 中国科学院半导体研究所 | Target ship detection method based on remote sensing image |
| CN110008833B (en) * | 2019-02-27 | 2021-03-26 | 中国科学院半导体研究所 | Target ship detection method based on optical remote sensing image |
| CN109871823A (en) * | 2019-03-11 | 2019-06-11 | 中国电子科技集团公司第五十四研究所 | A kind of satellite image Ship Detection of combination rotating frame and contextual information |
| CN109871823B (en) * | 2019-03-11 | 2021-08-31 | 中国电子科技集团公司第五十四研究所 | Satellite image ship detection method combining rotating frame and context information |
| CN110084181B (en) * | 2019-04-24 | 2021-04-20 | 哈尔滨工业大学 | Remote sensing image ship target detection method based on sparse MobileNet V2 network |
| CN110084181A (en) * | 2019-04-24 | 2019-08-02 | 哈尔滨工业大学 | A kind of remote sensing images Ship Target Detection method based on sparse MobileNetV2 network |
| CN111738236A (en) * | 2020-08-14 | 2020-10-02 | 之江实验室 | Image segmentation identification method, device and system of self-adaptive hierarchy |
| CN112766083A (en) * | 2020-12-30 | 2021-05-07 | 中南民族大学 | Remote sensing scene classification method and system based on multi-scale feature fusion |
| CN112766083B (en) * | 2020-12-30 | 2023-10-27 | 中南民族大学 | Remote sensing scene classification method and system based on multi-scale feature fusion |
| CN113256720A (en) * | 2021-06-03 | 2021-08-13 | 浙江大学 | Method for simultaneously detecting SAR image ship and trail thereof |
| CN116310734A (en) * | 2023-04-25 | 2023-06-23 | 慧铁科技有限公司 | Fault detection method and system for railway wagon running part based on deep learning |
| CN116310734B (en) * | 2023-04-25 | 2023-12-15 | 慧铁科技股份有限公司 | Fault detection method and system for railway wagon running part based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102867196B (en) | 2015-10-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102867196B (en) | Based on the complicated sea remote sensing image Ship Detection of Gist feature learning | |
| Nieto-Hidalgo et al. | Two-stage convolutional neural network for ship and spill detection using SLAR images | |
| CN110310264B (en) | DCNN-based large-scale target detection method and device | |
| CN109583369B (en) | Target identification method and device based on target area segmentation network | |
| Ao et al. | Detection and discrimination of ship targets in complex background from spaceborne ALOS-2 SAR images | |
| Zhang et al. | Ship detection for visual maritime surveillance from non-stationary platforms | |
| Corbane et al. | A complete processing chain for ship detection using optical satellite imagery | |
| CN105654091B (en) | Sea-surface target detection method and device | |
| Solberg et al. | Automatic detection of oil spills in ERS SAR images | |
| CN109427055B (en) | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy | |
| CN107247930A (en) | SAR image object detection method based on CNN and Selective Attention Mechanism | |
| CN109117802A (en) | Ship Detection towards large scene high score remote sensing image | |
| Capizzi et al. | A clustering based system for automated oil spill detection by satellite remote sensing | |
| CN102945378B (en) | Method for detecting potential target regions of remote sensing image on basis of monitoring method | |
| CN110414509B (en) | Port docking ship detection method based on sea-land segmentation and characteristic pyramid network | |
| CN110516605A (en) | Any direction Ship Target Detection method based on cascade neural network | |
| CN111079596A (en) | System and method for identifying typical marine artificial target of high-resolution remote sensing image | |
| Taravat et al. | Fully automatic dark-spot detection from SAR imagery with the combination of nonadaptive weibull multiplicative model and pulse-coupled neural networks | |
| CN109359661B (en) | Sentinel-1 radar image classification method based on convolutional neural network | |
| Hou et al. | SAR image ship detection based on visual attention model | |
| Hou et al. | End-to-end automatic ship detection and recognition in high-resolution Gaofen-3 spaceborne SAR images | |
| Corbane et al. | Fully automated procedure for ship detection using optical satellite imagery | |
| Zhao et al. | SAR ship detection based on end-to-end morphological feature pyramid network | |
| He et al. | Ship detection without sea-land segmentation for large-scale high-resolution optical satellite images | |
| CN105354824A (en) | Region extraction-based two-parameter constant false alarm detection method |
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 |