CN110472601A - A kind of Remote Sensing Target object identification method, device and storage medium - Google Patents
A kind of Remote Sensing Target object identification method, device and storage medium Download PDFInfo
- Publication number
- CN110472601A CN110472601A CN201910770755.2A CN201910770755A CN110472601A CN 110472601 A CN110472601 A CN 110472601A CN 201910770755 A CN201910770755 A CN 201910770755A CN 110472601 A CN110472601 A CN 110472601A
- Authority
- CN
- China
- Prior art keywords
- information
- candidate frame
- target
- target object
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/759—Region-based matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a kind of Remote Sensing Target object identification method, device and computer readable storage mediums, and the characteristic spectrum of image to be detected is calculated according to convolutional neural networks；Characteristic spectrum is input in region candidate frame network model, the corresponding characteristic information of each candidate frame is obtained；Judge whether each characteristic information matches with the characteristics of image of preset target object.When there is the target signature information with preset Image Feature Matching, using the coordinate information of candidate frame corresponding to linear regression algorithm adjustment target signature information, to obtain capture frame corresponding to target object in image to be detected.The coordinate information of target candidate frame is adjusted by dynamic, only need to generate a small amount of subgraph carries out classification judgement, can quickly recognize the position in image to be detected where target object, improve the recognition speed of target object.And by adjusting the coordinate information of target candidate frame, it can effectively improve the precision of target identification.
Description
Technical field
The present invention relates to satellite remote sensings and technical field of computer vision, more particularly to a kind of Remote Sensing Target object
Recognition methods, device and computer readable storage medium.
Background technique
Traditional remote sensing image target identification tool is based on sliding window detector, will be arranged in the picture different size of
Rectangle capture frame slides from left to right from top to bottom, whole image is traversed, so that many squares are obtained, successively in square
Image uses classifier, is input in convolutional neural networks and extracts the digitalized signature of image, then uses support vector machines
Scheduling algorithm is identified image category and is tightened capture frame using other linear regressor, until detection finishes, to sentence
Make whether the content in specified capture frame is specified target object.
Whole image is traversed using the method for sliding window detector, the subgraph quantity of generation is big, identifies the time of consuming
It is longer, cause the efficiency of images steganalysis lower.
It is those skilled in the art's problem to be solved as it can be seen that how to promote the recognition efficiency of image.
Summary of the invention
The purpose of the embodiment of the present invention is that providing a kind of Remote Sensing Target object identification method, device and computer-readable
Storage medium can promote the recognition efficiency of image.
In order to solve the above technical problems, the embodiment of the present invention provides a kind of Remote Sensing Target object identification method, comprising:
The characteristic spectrum of image to be detected is calculated according to convolutional neural networks；
The characteristic spectrum is input in region candidate frame network model, the corresponding characteristic information of each candidate frame is obtained；
Judge whether each characteristic information matches with the characteristics of image of preset target object；
When there is the target signature information with preset Image Feature Matching, then adjusted using linear regression algorithm
The coordinate information of candidate frame corresponding to the target signature information, to obtain in described image to be detected corresponding to target object
Capture frame.
Optionally, described that the characteristic spectrum is input in region candidate frame network model, it is corresponding to obtain each candidate frame
Characteristic information include:
The corresponding characteristic pattern of each candidate frame is extracted from the characteristic spectrum using region candidate frame network model；And root
The characteristic information of each characteristic pattern is extracted according to neural network.
Optionally, described when there is the target signature information with preset Image Feature Matching, then using linear
The coordinate information that regression algorithm adjusts candidate frame corresponding to the target signature information includes:
It is when there is the target signature information with preset Image Feature Matching, the target signature information institute is right
The position coordinates for the candidate frame answered are encoded into target feature vector；
According to the target feature vector and preset parameter information, loss function is calculated；Wherein, the parameter
Information includes the shift value and scale value of candidate frame；
Dynamic adjusts the value of the parameter information, until the loss function meets preset requirement, then according to adjustment after
Parameter information and target feature vector, determine the coordinate information of capture frame.
Optionally, it is obtained in described image to be detected after capture frame corresponding to target object described further include:
Show position of the capture frame in described image to be detected.
Optionally, the target object includes aircraft, steamer or warehouse.
The embodiment of the present invention also provides a kind of Remote Sensing Target object identification device, including computing unit, extraction unit,
Judging unit and adjustment unit；
The computing unit, for calculating the characteristic spectrum of image to be detected according to convolutional neural networks；
The extraction unit obtains each candidate for the characteristic spectrum to be input in region candidate frame network model
The corresponding characteristic information of frame；
The judging unit, for judge each characteristic information and preset target object characteristics of image whether
Matching；
The adjustment unit, it is for when existing and when the target signature information of preset Image Feature Matching, then sharp
With linear regression algorithm adjust the target signature information corresponding to candidate frame coordinate information, to obtain the mapping to be checked
The capture frame corresponding to target object as in.
Optionally, the extraction unit is specifically used for extracting from the characteristic spectrum using region candidate frame network model
The corresponding characteristic pattern of each candidate frame out；And the characteristic information of each characteristic pattern is extracted according to neural network.
Optionally, the adjustment unit includes coded sub-units, computation subunit and determining subelement；
The coded sub-units, for inciting somebody to action when the target signature information of presence and preset Image Feature Matching
The position coordinates of candidate frame corresponding to the target signature information are encoded into target feature vector；
The computation subunit, for calculating damage according to the target feature vector and preset parameter information
Lose function；Wherein, the parameter information includes the shift value and scale value of candidate frame；
The determining subelement, for dynamically adjusting the value of the parameter information, until the loss function meets in advance
If it is required that then determining the coordinate information of capture frame according to parameter information adjusted and target feature vector.
It optionally, further include display unit；
The display unit, for it is described obtain in described image to be detected capture frame corresponding to target object it
Afterwards, position of the capture frame in described image to be detected is shown.
Optionally, the target object includes aircraft, steamer or warehouse.
The embodiment of the invention also provides a kind of Remote Sensing Target object identification devices, comprising:
Memory, for storing computer program；
Processor, for executing the computer program to realize the Remote Sensing Target object as described in above-mentioned any one
The step of recognition methods.
The embodiment of the invention also provides a kind of computer readable storage medium, deposited on the computer readable storage medium
Computer program is contained, the Remote Sensing Target object as described in any of the above-described is realized when the computer program is executed by processor
Body recognition methods step.
The characteristic spectrum of image to be detected is calculated according to convolutional neural networks it can be seen from above-mentioned technical proposal；It will be special
Sign map is input in region candidate frame network model, obtains the corresponding characteristic information of each candidate frame；Judge each characteristic information with
Whether the characteristics of image of preset target object matches.When in the presence of the target signature with preset Image Feature Matching
When information, then illustrate that in the corresponding target candidate frame of the target signature information include target object, in order to enable target candidate
Frame preferably selectes target object, can use the seat of candidate frame corresponding to linear regression algorithm adjustment target signature information
Information is marked, to obtain capture frame corresponding to target object in image to be detected.The coordinate of target candidate frame is adjusted by dynamic
Information, only need to generate a small amount of subgraph carries out classification judgement, can quickly recognize in image to be detected where target object
Position, improve the recognition speed of target object.And by adjusting the coordinate information of target candidate frame, thus it is possible to vary candidate
The size and location of frame effectively raises the precision of target identification.
Detailed description of the invention
In order to illustrate the embodiments of the present invention more clearly, attached drawing needed in the embodiment will be done simply below
It introduces, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ordinary skill people
For member, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of Remote Sensing Target object identification method provided in an embodiment of the present invention；
Fig. 2 a be one provided in an embodiment of the present invention include multi-aircraft image；
Fig. 2 b is a kind of target object place that adjustment candidate frame coordinate information obtains later provided in an embodiment of the present invention
The schematic diagram of the capture frame of position；
Fig. 3 is a kind of structural schematic diagram of Remote Sensing Target object identification device provided in an embodiment of the present invention；
Fig. 4 is a kind of hardware structural diagram of Remote Sensing Target object identification device provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, rather than whole embodiments.Based on this
Embodiment in invention, those of ordinary skill in the art are without making creative work, obtained every other
Embodiment belongs to the scope of the present invention.
In order to enable those skilled in the art to better understand the solution of the present invention, with reference to the accompanying drawings and detailed description
The present invention is described in further detail.
Next, a kind of Remote Sensing Target object identification method provided by the embodiment of the present invention is discussed in detail.Fig. 1 is
A kind of flow chart of Remote Sensing Target object identification method provided in an embodiment of the present invention, this method comprises:
S101: the characteristic spectrum of image to be detected is calculated according to convolutional neural networks.
Because the Pixel Information data volume that image to be detected itself includes is very big, if being used directly to calculate will occupy greatly
The storage and computing resource of amount, therefore in embodiments of the present invention, the characteristic spectrum of image to be detected can be first extracted, further
Characteristic information is obtained, in order to carry out subsequent calculating.
Characteristic spectrum is equivalent to a kind of mathematic(al) manipulation of original picture, and the information of picture can be indicated with the mode more simplified.
S102: characteristic spectrum is input in region candidate frame network model, obtains the corresponding characteristic information of each candidate frame.
Be preset with multiple candidate frames in region candidate frame network model, under original state, each candidate frame it is big
Small and position is fixed.
In practical applications, it can use region candidate frame network model and extract each candidate frame correspondence from characteristic spectrum
Characteristic pattern.For the ease of subsequent analysis, the characteristic information of each characteristic pattern can be extracted according to neural network.
S103: judge whether each characteristic information matches with the characteristics of image of preset target object.
The object that target object identifies needed for referring in image, target object can be aircraft, steamer or warehouse etc..
Different types of target object, corresponding characteristics of image are different.It in embodiments of the present invention, can be preparatory
The image for including target object is handled, the corresponding characteristic information in target object position is extracted, by big
Amount includes that the image pattern of target object is trained, to obtain the characteristics of image of target object.
When the Image Feature Matching of the corresponding characteristic information of some candidate frame and target object, then illustrate in the candidate frame
It include target object.Since the size and location of candidate frame under original state is relatively fixed, the candidate frame is possible can not be complete
Entirely include target object, needs further to adjust the size and location of candidate frame at this time, so that candidate frame can be more quasi-
True frame selects target object.Therefore, it when there is the target signature information with preset Image Feature Matching, then executes
S104。
S104: using the coordinate information of candidate frame corresponding to linear regression algorithm adjustment target signature information, to obtain
Capture frame corresponding to target object in image to be detected.
Candidate frame position in the picture and size can be determined according to coordinate information.In embodiments of the present invention, will
Candidate frame after adjustment coordinate is referred to as capture frame.
For the ease of the calculating of coordinate information, in embodiments of the present invention, need to convert coordinate information to the shape of vector
Formula.
Specifically, when there is the target signature information with preset Image Feature Matching, by target signature information
The position coordinates of corresponding candidate frame are encoded into target feature vector.According to target feature vector and preset parameter
Information calculates loss function.
Wherein, parameter information includes the shift value and scale value of candidate frame.
Under original state, the position coordinates of candidate frame are Given information, by adjusting the parameter information of candidate frame, Ke Yigai
Become position and the size of candidate frame.
In embodiments of the present invention, the value of dynamically-adjusting parameter information is carried out according to loss function, until loss function is full
Sufficient preset requirement determines the coordinate information of capture frame then according to parameter information adjusted and target feature vector.
For example, passing through translation and scaling gradually approaching to reality window for original candidate frame.When calculating, use
The method of linear regression gives the feature vector, X of the coordinate of original window, learns one group of parameter W, is made by calculating loss function
W is continued to optimize, and when loss function meets preset requirement, then illustrates that the position of candidate frame at this time is optimal, finally obtained WX is
For the coordinate information of capture frame.
In embodiments of the present invention, in obtaining image to be detected after capture frame corresponding to target object, Ke Yizhan
Show position of the capture frame in image to be detected, in order to which user can intuitively check target object in image to be detected
Distribution situation.
It as shown in Figure 2 a include the image of multi-aircraft for one, aircraft is the target object of required identification, such as Fig. 2 b
It is shown the schematic diagram of the capture frame of the target object position obtained after adjustment candidate frame coordinate information, Fig. 2 b can be true
7 capture frames are made, the size and location of this 7 capture frames is different.
Traditional target identification etc. cannot be locked accurately than zooming in or out capture frame, while equidistant mobile capture frame
The accurate location and size of capture frame, recognition speed is slow, and error is larger.Use remote sensing images provided in an embodiment of the present invention
Target object recognition methods is constantly adjusted position and the size of candidate frame based on characteristics of image, can effectively improve target identification
Speed and precision.
The characteristic spectrum of image to be detected is calculated according to convolutional neural networks it can be seen from above-mentioned technical proposal；It will be special
Sign map is input in region candidate frame network model, obtains the corresponding characteristic information of each candidate frame；Judge each characteristic information with
Whether the characteristics of image of preset target object matches.When in the presence of the target signature with preset Image Feature Matching
When information, then illustrate that in the corresponding target candidate frame of the target signature information include target object, in order to enable target candidate
Frame preferably selectes target object, can use the seat of candidate frame corresponding to linear regression algorithm adjustment target signature information
Information is marked, to obtain capture frame corresponding to target object in image to be detected.The coordinate of target candidate frame is adjusted by dynamic
Information, only need to generate a small amount of subgraph carries out classification judgement, can quickly recognize in image to be detected where target object
Position, improve the recognition speed of target object.And by adjusting the coordinate information of target candidate frame, thus it is possible to vary candidate
The size and location of frame effectively raises the precision of target identification.
Fig. 3 is a kind of structural schematic diagram of Remote Sensing Target object identification device provided in an embodiment of the present invention, including
Computing unit 31, extraction unit 32, judging unit 33 and adjustment unit 34；
Computing unit 31, for calculating the characteristic spectrum of image to be detected according to convolutional neural networks；
It is corresponding to obtain each candidate frame for characteristic spectrum to be input in region candidate frame network model for extraction unit 32
Characteristic information；
Judging unit 33, for judging whether each characteristic information matches with the characteristics of image of preset target object；
Adjustment unit 34, for when there is the target signature information with preset Image Feature Matching, then utilizing
Linear regression algorithm adjusts the coordinate information of candidate frame corresponding to target signature information, to obtain object in image to be detected
Capture frame corresponding to body.
Optionally, extraction unit is specifically used for extracting each candidate from characteristic spectrum using region candidate frame network model
The corresponding characteristic pattern of frame；And the characteristic information of each characteristic pattern is extracted according to neural network.
Optionally, adjustment unit includes coded sub-units, computation subunit and determining subelement；
Coded sub-units, for when exist and preset Image Feature Matching target signature information when, by target
The position coordinates of candidate frame corresponding to characteristic information are encoded into target feature vector；
Computation subunit, for calculating loss function according to target feature vector and preset parameter information；Its
In, parameter information includes the shift value and scale value of candidate frame；
Determine subelement, for the value of dynamically-adjusting parameter information, until loss function meets preset requirement, then basis
Parameter information and target feature vector adjusted, determine the coordinate information of capture frame.
It optionally, further include display unit；
Display unit shows capture frame after capture frame corresponding to the target object in obtaining image to be detected
Position in image to be detected.
Optionally, target object includes aircraft, steamer or warehouse.
The explanation of feature may refer to the related description of embodiment corresponding to Fig. 1 in embodiment corresponding to Fig. 3, here no longer
It repeats one by one.
The characteristic spectrum of image to be detected is calculated according to convolutional neural networks it can be seen from above-mentioned technical proposal；It will be special
Sign map is input in region candidate frame network model, obtains the corresponding characteristic information of each candidate frame；Judge each characteristic information with
Whether the characteristics of image of preset target object matches.When in the presence of the target signature with preset Image Feature Matching
When information, then illustrate that in the corresponding target candidate frame of the target signature information include target object, in order to enable target candidate
Frame preferably selectes target object, can use the seat of candidate frame corresponding to linear regression algorithm adjustment target signature information
Information is marked, to obtain capture frame corresponding to target object in image to be detected.The coordinate of target candidate frame is adjusted by dynamic
Information, only need to generate a small amount of subgraph carries out classification judgement, can quickly recognize in image to be detected where target object
Position, improve the recognition speed of target object.And by adjusting the coordinate information of target candidate frame, thus it is possible to vary candidate
The size and location of frame effectively raises the precision of target identification.
Fig. 4 is a kind of hardware configuration signal of Remote Sensing Target object identification device 40 provided in an embodiment of the present invention
Figure, comprising:
Memory 41, for storing computer program；
Processor 42, for executing computer program to realize such as above-mentioned any one Remote Sensing Target object identification side
The step of method.
The embodiment of the invention also provides a kind of computer readable storage medium, it is stored on computer readable storage medium
Computer program realizes that any of the above-described Remote Sensing Target object identification method such as walks when computer program is executed by processor
Suddenly.
Being provided for the embodiments of the invention a kind of Remote Sensing Target object identification method, device and computer above can
Storage medium is read to be described in detail.Each embodiment is described in a progressive manner in specification, each embodiment emphasis
What is illustrated is the difference from other embodiments, and the same or similar parts in each embodiment may refer to each other.For
For device disclosed in embodiment, since it is corresponded to the methods disclosed in the examples, so be described relatively simple, correlation
Place is referring to method part illustration.It should be pointed out that for those skilled in the art, not departing from this hair
, can be with several improvements and modifications are made to the present invention under the premise of bright principle, these improvement and modification also fall into power of the present invention
In the protection scope that benefit requires.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think beyond the scope of this invention.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
Claims (10)
1. a kind of Remote Sensing Target object identification method characterized by comprising
The characteristic spectrum of image to be detected is calculated according to convolutional neural networks；
The characteristic spectrum is input in region candidate frame network model, the corresponding characteristic information of each candidate frame is obtained；
Judge whether each characteristic information matches with the characteristics of image of preset target object；
When there is the target signature information with preset Image Feature Matching, then using described in linear regression algorithm adjustment
The coordinate information of candidate frame corresponding to target signature information, to obtain taking corresponding to target object in described image to be detected
Frame.
2. the method according to claim 1, wherein described be input to region candidate frame net for the characteristic spectrum
In network model, obtaining the corresponding characteristic information of each candidate frame includes:
The corresponding characteristic pattern of each candidate frame is extracted from the characteristic spectrum using region candidate frame network model；And according to mind
The characteristic information of each characteristic pattern is extracted through network.
3. the method according to claim 1, wherein it is described when exist and preset Image Feature Matching
When target signature information, then the coordinate information of candidate frame corresponding to the target signature information is adjusted using linear regression algorithm
Include:
It, will be corresponding to the target signature information when there is the target signature information with preset Image Feature Matching
The position coordinates of candidate frame are encoded into target feature vector；
According to the target feature vector and preset parameter information, loss function is calculated；Wherein, the parameter information
Shift value and scale value including candidate frame；
Dynamic adjusts the value of the parameter information, until the loss function meets preset requirement, then according to ginseng adjusted
Number information and target feature vector, determine the coordinate information of capture frame.
4. the method according to claim 1, wherein obtaining target object institute in described image to be detected described
After corresponding capture frame further include:
Show position of the capture frame in described image to be detected.
5. method according to any of claims 1-4, which is characterized in that the target object includes aircraft, steamer
Or warehouse.
6. a kind of Remote Sensing Target object identification device, which is characterized in that including computing unit, extraction unit, judging unit
And adjustment unit；
The computing unit, for calculating the characteristic spectrum of image to be detected according to convolutional neural networks；
The extraction unit obtains each candidate frame pair for the characteristic spectrum to be input in region candidate frame network model
The characteristic information answered；
The judging unit, for judge each characteristic information and preset target object characteristics of image whether
Match；
The adjustment unit, for when there is the target signature information with preset Image Feature Matching, then utilizing line
Property regression algorithm adjusts the coordinate information of candidate frame corresponding to the target signature information, to obtain in described image to be detected
Capture frame corresponding to target object.
7. device according to claim 6, which is characterized in that the extraction unit is specifically used for utilizing region candidate frame net
Network model extracts the corresponding characteristic pattern of each candidate frame from the characteristic spectrum；And each characteristic pattern is extracted according to neural network
Characteristic information.
8. device according to claim 6, which is characterized in that the adjustment unit includes coded sub-units, calculates son list
Member and determining subelement；
The coded sub-units will be described for when existing and when the target signature information of preset Image Feature Matching
The position coordinates of candidate frame corresponding to target signature information are encoded into target feature vector；
The computation subunit, for calculating loss letter according to the target feature vector and preset parameter information
Number；Wherein, the parameter information includes the shift value and scale value of candidate frame；
The determining subelement, for dynamically adjusting the value of the parameter information, until the loss function meets default want
It asks, then according to parameter information adjusted and target feature vector, determines the coordinate information of capture frame.
9. a kind of Remote Sensing Target object identification device characterized by comprising
Memory, for storing computer program；
Processor, for executing the computer program to realize the Remote Sensing Target as described in claim 1 to 5 any one
The step of object identification method.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program realizes the Remote Sensing Target object as described in any one of claim 1 to 5 when the computer program is executed by processor
Recognition methods step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910770755.2A CN110472601B (en) | 2019-08-20 | 2019-08-20 | Remote sensing image target object identification method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910770755.2A CN110472601B (en) | 2019-08-20 | 2019-08-20 | Remote sensing image target object identification method, device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110472601A true CN110472601A (en) | 2019-11-19 |
| CN110472601B CN110472601B (en) | 2023-05-02 |
Family
ID=68513058
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910770755.2A Active CN110472601B (en) | 2019-08-20 | 2019-08-20 | Remote sensing image target object identification method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110472601B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610113A (en) * | 2017-09-13 | 2018-01-19 | 北京邮电大学 | The detection method and device of Small object based on deep learning in a kind of image |
| CN108416378A (en) * | 2018-02-28 | 2018-08-17 | 电子科技大学 | A kind of large scene SAR target identification methods based on deep neural network |
| CN108921875A (en) * | 2018-07-09 | 2018-11-30 | 哈尔滨工业大学（深圳） | A kind of real-time traffic flow detection and method for tracing based on data of taking photo by plane |
| CN109101897A (en) * | 2018-07-20 | 2018-12-28 | 中国科学院自动化研究所 | Object detection method, system and the relevant device of underwater robot |
| US20190087673A1 (en) * | 2017-09-15 | 2019-03-21 | Baidu Online Network Technology (Beijing) Co., Ltd | Method and apparatus for identifying traffic light |
| CN110084093A (en) * | 2019-02-20 | 2019-08-02 | 北京航空航天大学 | The method and device of object detection and recognition in remote sensing images based on deep learning |
-
2019
- 2019-08-20 CN CN201910770755.2A patent/CN110472601B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107610113A (en) * | 2017-09-13 | 2018-01-19 | 北京邮电大学 | The detection method and device of Small object based on deep learning in a kind of image |
| US20190087673A1 (en) * | 2017-09-15 | 2019-03-21 | Baidu Online Network Technology (Beijing) Co., Ltd | Method and apparatus for identifying traffic light |
| CN108416378A (en) * | 2018-02-28 | 2018-08-17 | 电子科技大学 | A kind of large scene SAR target identification methods based on deep neural network |
| CN108921875A (en) * | 2018-07-09 | 2018-11-30 | 哈尔滨工业大学（深圳） | A kind of real-time traffic flow detection and method for tracing based on data of taking photo by plane |
| CN109101897A (en) * | 2018-07-20 | 2018-12-28 | 中国科学院自动化研究所 | Object detection method, system and the relevant device of underwater robot |
| CN110084093A (en) * | 2019-02-20 | 2019-08-02 | 北京航空航天大学 | The method and device of object detection and recognition in remote sensing images based on deep learning |
Non-Patent Citations (7)
| Title |
|---|
| JINSU LEE ET AL.: "Object Detection With Sliding Window in Images Including Multiple Similar Objects" * |
| 刘建国等: "基于YOLOv2的行人检测方法研究", 《数字制造科学》 * |
| 唐小佩等: "遥感图像飞机目标候选区域选取方法研究", 《计算机仿真》 * |
| 喻一凡等: "线性回归的渡船精准定位方法研究", 《小型微型计算机系统》 * |
| 贾瑞: "基于RCNN的车辆区域检测研究", 《数字技术与应用》 * |
| 郭智等: "基于深度卷积神经网络的遥感图像飞机目标检测方法", 《电子与信息学报》 * |
| 龚安等: "基于卷积神经网络的实时行人检测方法", 《计算机系统应用》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110472601B (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163198B (en) | Table identification reconstruction method and device and storage medium | |
| CN106056064B (en) | A kind of face identification method and face identification device | |
| CN105493078B (en) | Colored sketches picture search | |
| US8265351B2 (en) | Method, system and computer program product for automatic and semi-automatic modification of digital images of faces | |
| US8620038B2 (en) | Method, system and computer program product for automatic and semi-automatic modification of digital images of faces | |
| CN106030661B (en) | The independent 3D scene texture background in the visual field | |
| US8660319B2 (en) | Method, system and computer program product for automatic and semi-automatic modification of digital images of faces | |
| CN103810490B (en) | A kind of method and apparatus for the attribute for determining facial image | |
| CN109255324A (en) | Gesture processing method, interaction control method and equipment | |
| CN105160323B (en) | A kind of gesture identification method | |
| CN111008935B (en) | Face image enhancement method, device, system and storage medium | |
| CN111401318B (en) | Action recognition method and device | |
| CN109598234A (en) | Critical point detection method and apparatus | |
| CN105139007B (en) | Man face characteristic point positioning method and device | |
| CN102567716A (en) | Face synthetic system and implementation method | |
| CN105069444B (en) | A kind of gesture identifying device | |
| CN110059680A (en) | A kind of detection method of ID Card Image, device and equipment | |
| CN109948521A (en) | Image correcting error method and device, equipment and storage medium | |
| CN111860091A (en) | Face image evaluation method and system, server and computer readable storage medium | |
| CN105205482A (en) | Quick facial feature recognition and posture estimation method | |
| CN106371614A (en) | Gesture recognition optimizing method and device | |
| CN110007764B (en) | Gesture skeleton recognition method, device and system and storage medium | |
| KR20160046399A (en) | Method and Apparatus for Generation Texture Map, and Database Generation Method | |
| CN109919128A (en) | Acquisition methods, device and the electronic equipment of control instruction | |
| CN106933976A (en) | Set up the method for the net models of human body 3D and its application in 3D fittings |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |