CN111126189A - Target searching method based on remote sensing image - Google Patents
Target searching method based on remote sensing image Download PDFInfo
- Publication number
- CN111126189A CN111126189A CN201911258224.1A CN201911258224A CN111126189A CN 111126189 A CN111126189 A CN 111126189A CN 201911258224 A CN201911258224 A CN 201911258224A CN 111126189 A CN111126189 A CN 111126189A
- Authority
- CN
- China
- Prior art keywords
- information
- remote sensing
- array
- acquiring
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 claims abstract description 23
- 238000004458 analytical method Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims abstract description 12
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 9
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 9
- 238000005070 sampling Methods 0.000 claims abstract description 5
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 239000002131 composite material Substances 0.000 claims description 2
- 230000008569 process Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000011176 pooling Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 229940088594 vitamin Drugs 0.000 description 1
- 229930003231 vitamin Natural products 0.000 description 1
- 235000013343 vitamin Nutrition 0.000 description 1
- 239000011782 vitamin Substances 0.000 description 1
- 150000003722 vitamin derivatives Chemical class 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a target searching method based on remote sensing images, which comprises the following steps: acquiring original remote sensing image information, and decomposing an original remote sensing image into overlapped picture slices; analyzing and acquiring picture slice information, and establishing a distributed non-relational database based on the picture slice information; inserting the picture slicing information into a distributed non-relational database; sending the picture slice information to a neural network, acquiring and processing the picture slice information by the neural network, and outputting picture slice analysis result information; acquiring picture slice analysis result information, storing the picture slice analysis result information to a new array, processing element information in the new array, and outputting a synthesis matrix; adopting a maximum pool algorithm to perform down-sampling synthesis on a maximum matrix; and identifying the image and acquiring a matched image. By adopting the method, on one hand, the accuracy rate of searching the remote sensing image is improved, and on the other hand, the efficiency of searching the remote sensing image is improved.
Description
Technical Field
The invention relates to the technical field of remote sensing image target searching, in particular to a target searching method based on remote sensing images.
Background
It is not an easy matter to automatically detect objects from remote sensing images, and automatic object recognition algorithms must be able to recognize three-dimensional objects whose exact shape may not be well understood, and which may appear in any direction where lighting and visibility conditions vary widely.
In the traditional automatic target identification method, a target is separated from a surrounding area by extracting a contour, and then the shape of the target is identified according to the described characteristics.
In fact, it is very difficult to obtain a real contour from an image, the error in contour restoration reduces the probability of accurately detecting a target, and the problem is complicated by various factors such as illumination conditions, viewing angles, atmospheric conditions, and the like.
Disclosure of Invention
Aiming at the technical problems in the related art, the invention provides a target searching method based on remote sensing images, which can overcome the defects in the prior art.
In order to achieve the technical purpose, the technical scheme of the invention is realized as follows:
a target searching method based on remote sensing images comprises the following steps:
s1: acquiring original remote sensing image information, and decomposing an original remote sensing image into overlapped picture slices;
s2: analyzing and acquiring picture slice information, and establishing a distributed non-relational database based on the picture slice information;
s3: inserting the picture slicing information into a distributed non-relational database;
s4: sending the picture slice information to a neural network, acquiring and processing the picture slice information by the neural network, and outputting picture slice analysis result information;
s5: acquiring picture slice analysis result information, storing the picture slice analysis result information to a new array, processing element information in the new array, and outputting a synthesis matrix;
s6: adopting a maximum pool algorithm to perform down-sampling synthesis on a maximum matrix;
s7: and identifying the image and acquiring a matched image.
Further, the step S1 includes the following steps:
s11: establishing a sliding window to obtain an original remote sensing image;
s12: decomposing the original remote sensing image into a plurality of image tiles with the same size;
s13: and acquiring and storing image tile information.
Further, the step of acquiring and processing the picture slice information by the neural network in step S4 includes the following steps:
s41: acquiring image tile information;
s42: analyzing and calculating the weight of the image tile;
s43: and finding and marking the target object.
Further, the step of saving the picture slice analysis result information to the new array in step S5, processing the element information in the new array, and outputting the synthesis matrix includes the following steps:
s51: according to the original tile layout format, storing the picture slicing result to a grid, and synthesizing an array;
s52: obtaining original remote sensing image marking result information;
s53: and outputting the array.
Further, the step S6 includes the following steps:
s61: setting the size specification of the array, and acquiring the information of the marked array;
s62: dividing the marked array into grids according to the size specification of the array;
s63: analyzing the segmented grid data to obtain a maximum grid;
s64: add the maximum grid to the maximum pool array.
Further, the step S7 includes the following steps:
s71: acquiring data information in the maximum pool array, and sending the data information in the maximum pool array to a new neural network;
s72: identifying a new neural network by full connection;
s73: and judging the image matching degree and outputting a result.
The invention has the beneficial effects that: by adopting the method, when the same target is identified under the same conditions of the same original image, the same computer hardware equipment, the same network environment and the like, and the accuracy rate reaches 95%, the time consumed by adopting the patent technology is one tenth of that of the traditional method, namely the efficiency is improved by 9 times. Therefore, on one hand, the accuracy rate of searching the remote sensing image is improved, and on the other hand, the efficiency of searching the remote sensing image is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a flowchart illustrating steps of a method for searching a target based on a remote sensing image according to an embodiment of the present invention.
Fig. 2 is an actual case recognition target diagram of a target searching method based on remote sensing images according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present invention.
As shown in fig. 1, according to the method for searching for a target based on a remote sensing image in the embodiment of the present invention, S1: acquiring original remote sensing image information, and decomposing an original remote sensing image into overlapped picture slices;
s2: analyzing and acquiring picture slice information, and establishing a distributed non-relational database based on the picture slice information;
s3: inserting the picture slicing information into a distributed non-relational database;
s4: sending the picture slice information to a neural network, acquiring and processing the picture slice information by the neural network, and outputting picture slice analysis result information;
s5: acquiring picture slice analysis result information, storing the picture slice analysis result information to a new array, processing element information in the new array, and outputting a synthesis matrix;
s6: adopting a maximum pool algorithm to perform down-sampling synthesis on a maximum matrix;
s7: and identifying the image and acquiring a matched image.
Step S1 includes the following steps:
s11: establishing a sliding window to obtain an original remote sensing image;
s12: decomposing the original remote sensing image into a plurality of image tiles with the same size;
s13: and acquiring and storing image tile information.
The step of acquiring and processing the picture slice information by the neural network in step S4 includes the following steps:
s41: acquiring image tile information;
s42: analyzing and calculating the weight of the image tile;
s43: and finding and marking the target object.
In step S5, the step of saving the picture slice analysis result information to the new array, processing the element information in the new array, and outputting the composite matrix includes the following steps:
s51: according to the original tile layout format, storing the picture slicing result to a grid, and synthesizing an array;
s52: obtaining original remote sensing image marking result information;
s53: and outputting the array.
Step S6 includes the following steps:
s61: setting the size specification of the array, and acquiring the information of the marked array;
s62: dividing the marked array into grids according to the size specification of the array;
s63: analyzing the segmented grid data to obtain a maximum grid;
s64: add the maximum grid to the maximum pool array.
Step S7 includes the following steps:
s71: acquiring data information in the maximum pool array, and sending the data information in the maximum pool array to a new neural network;
s72: identifying a new neural network by full connection;
s73: and judging the image matching degree and outputting a result.
In order to facilitate understanding of the above-described technical aspects of the present invention, the above-described technical aspects of the present invention will be described in detail below in terms of specific usage.
The invention adopts the following points:
1) improved polynomial regression
The nonlinear problem can be treated by improved polynomial regression, and any function can be approximated by a polynomial in a segmentation way, and the improved polynomial regression can be used for analysis no matter the relation between a variable and other independent variables.
Let x1＝x，x2＝x2，…，xm＝xm
Then
2) Clustering
The goal of clustering is to group observations with similar characteristics.
Assuming that data { Xi } needs to be aggregated into k classes, the class to which each data belongs after clustering is { Ti }, and the centers of the k clusters are { mu i }. The function is then:
3) reducing vitamin
Dimensionality reduction is the removal of the least important information (sometimes redundant columns) from the dataset. In practice, it is often seen that there are hundreds or even thousands of columns (also called elements) of data sets, so it is crucial to reduce the total number. For example, an image may contain thousands of pixels, not all of which are important to analysis, many of which provide redundant information. In these cases, a dimensionality reduction algorithm is required to make the data set easy to manage.
Suppose that the high-dimensional data is represented by X, Xi represents the ith sample, the low-dimensional data is represented by Y, and Yi represents the ith sample.
Then the process of the first step is carried out,
4) integration
The integrated approach combines multiple predictive models to obtain a higher quality prediction than each model provides individually. The integration method is considered as one method of reducing the variance and bias of a single machine learning model. As any given model may be accurate in some cases, but not in others. In another model, the relative accuracy may be reversed. By combining the two models, the quality of the prediction can be balanced.
Assuming the basis functions are:
then, the integrated process result may be
In particular use, a process called convolution is required to make the neural network understand the translation invariance. The idea of convolution is motivated in part by computer science and in part by biology.
Step 1: decomposing an image into overlapping image slices
Similar to the sliding window search above, the original image is converted into small image tiles of the same size by using a sliding window over the entire original image and saving each result as a separate small picture (i.e., tile).
Step 2: inputting each image tile into a small neural network
As previously done, each individual image tile is input to the neural network as an image, and this process is performed once for each tile. Among these, there is a very different setup: the same neural network weights are reserved for each individual tile in the same original image. In other words, each picture is being treated equally. If an object appears in any given tile, that tile will be marked.
And 3, step 3: saving the results of each tile to a new array
The layout of the original tile cannot be discarded. Thus, the result of processing each tile is saved as a grid arranged identically to the original image. To this end, a large image is input and processed to obtain a slightly smaller array that records which portions of the original image are marked.
And 4, step 4: down sampling
The result of step 3 is an array that maps which portions of the original image are marked. However, this array is still large and needs to be downsampled using the max-pool algorithm in order to reduce the size of the array. The tag array is divided into grids according to a certain size, and only the most important tags are reserved in each grid. This not only reduces the size of the array, but also retains the most important labels.
And 5, step 5: image recognition
To date, a large image has been reduced from the original to a relatively small array, but the matrix is still a pile of numbers, and in order to be able to identify the target object, the set of data needs to be used as an input for another neural network. This final neural network will decide whether the images match.
The above is the complete process of image recognition, and the image processing pipeline is a series of steps: convolution, max pooling, and finally a fully connected identification network.
When solving real-world problems, the steps can be combined and stacked as many times as needed, and one process can have two, three or more convolution layers, and the more convolution steps are possessed, the more complex functions the network can learn to identify.
As shown in fig. 2, one practical application case applies convolution and max pooling 2 times, then convolution 3 times, max pooling 3 times, and then joins two fully joined layers for an original image of 224x 224 pixels. The end result is that objects in the image can be identified from 1000 categories.
In summary, with the above technical solution of the present invention, when the same target is identified by using the method under the same conditions of the same original image, the same computer hardware device, the same network environment, and the like, and the accuracy rate reaches 95%, the time consumed by using the technology of this patent is one tenth of that of the conventional method, that is, the efficiency is improved by 9 times. Therefore, on one hand, the accuracy rate of searching the remote sensing image is improved, and on the other hand, the efficiency of searching the remote sensing image is improved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (6)
1. A target searching method based on remote sensing images is characterized by comprising the following steps:
s1: acquiring original remote sensing image information, and decomposing an original remote sensing image into overlapped picture slices;
s2: analyzing and acquiring picture slice information, and establishing a distributed non-relational database based on the picture slice information;
s3: inserting the picture slicing information into a distributed non-relational database;
s4: sending the picture slice information to a neural network, acquiring and processing the picture slice information by the neural network, and outputting picture slice analysis result information;
s5: acquiring picture slice analysis result information, storing the picture slice analysis result information to a new array, processing element information in the new array, and outputting a synthesis matrix;
s6: adopting a maximum pool algorithm to perform down-sampling synthesis on a maximum matrix;
s7: and identifying the image and acquiring a matched image.
2. The method for finding a target based on a remote sensing image according to claim 1, wherein the step S1 comprises the following steps:
s11: establishing a sliding window to obtain an original remote sensing image;
s12: decomposing the original remote sensing image into a plurality of image tiles with the same size;
s13: and acquiring and storing image tile information.
3. The method for finding a target based on remote sensing images according to claim 1, wherein the step of obtaining and processing the picture slice information by the neural network in the step S4 comprises the following steps:
s41: acquiring image tile information;
s42: analyzing and calculating the weight of the image tile;
s43: and finding and marking the target object.
4. The method for finding a target based on a remote sensing image according to claim 1, wherein the step of saving the picture slice analysis result information to a new array in the step S5, processing element information in the new array, and outputting a composite matrix comprises the following steps:
s51: according to the original tile layout format, storing the picture slicing result to a grid, and synthesizing an array;
s52: obtaining original remote sensing image marking result information;
s53: and outputting the array.
5. The method for finding a target based on a remote sensing image according to claim 1, wherein the step S6 comprises the following steps:
s61: setting the size specification of the array, and acquiring the information of the marked array;
s62: dividing the marked array into grids according to the size specification of the array;
s63: analyzing the segmented grid data to obtain a maximum grid;
s64: add the maximum grid to the maximum pool array.
6. The method for finding a target based on a remote sensing image according to claim 1, wherein the step S7 comprises the following steps:
s71: acquiring data information in the maximum pool array, and sending the data information in the maximum pool array to a new neural network;
s72: identifying a new neural network by full connection;
s73: and judging the image matching degree and outputting a result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911258224.1A CN111126189A (en) | 2019-12-10 | 2019-12-10 | Target searching method based on remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911258224.1A CN111126189A (en) | 2019-12-10 | 2019-12-10 | Target searching method based on remote sensing image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111126189A true CN111126189A (en) | 2020-05-08 |
Family
ID=70498070
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911258224.1A Pending CN111126189A (en) | 2019-12-10 | 2019-12-10 | Target searching method based on remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111126189A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020325A (en) * | 2013-01-17 | 2013-04-03 | 中国科学院计算机网络信息中心 | Distributed remote sensing data organization query method based on NoSQL database |
| CN107330405A (en) * | 2017-06-30 | 2017-11-07 | 上海海事大学 | Remote sensing images Aircraft Target Recognition based on convolutional neural networks |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN108932303A (en) * | 2018-06-12 | 2018-12-04 | 中国电子科技集团公司第二十八研究所 | A kind of distribution visual remote sensing image Detection dynamic target and analysis system |
| CN109117802A (en) * | 2018-08-21 | 2019-01-01 | 东北大学 | Ship Detection towards large scene high score remote sensing image |
| CN109583369A (en) * | 2018-11-29 | 2019-04-05 | 北京邮电大学 | A kind of target identification method and device based on target area segmentation network |
-
2019
- 2019-12-10 CN CN201911258224.1A patent/CN111126189A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020325A (en) * | 2013-01-17 | 2013-04-03 | 中国科学院计算机网络信息中心 | Distributed remote sensing data organization query method based on NoSQL database |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN107330405A (en) * | 2017-06-30 | 2017-11-07 | 上海海事大学 | Remote sensing images Aircraft Target Recognition based on convolutional neural networks |
| CN108932303A (en) * | 2018-06-12 | 2018-12-04 | 中国电子科技集团公司第二十八研究所 | A kind of distribution visual remote sensing image Detection dynamic target and analysis system |
| CN109117802A (en) * | 2018-08-21 | 2019-01-01 | 东北大学 | Ship Detection towards large scene high score remote sensing image |
| CN109583369A (en) * | 2018-11-29 | 2019-04-05 | 北京邮电大学 | A kind of target identification method and device based on target area segmentation network |
Non-Patent Citations (1)
| Title |
|---|
| 季顺平 等: "遥感影像建筑物提取的卷积神经元网络与开源数据集方法" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bosilj et al. | Transfer learning between crop types for semantic segmentation of crops versus weeds in precision agriculture | |
| US11681418B2 (en) | Multi-sample whole slide image processing in digital pathology via multi-resolution registration and machine learning | |
| Lin et al. | Automated defect inspection of LED chip using deep convolutional neural network | |
| Kumar et al. | Leafsnap: A computer vision system for automatic plant species identification | |
| CN103400143B (en) | A kind of data Subspace clustering method based on various visual angles | |
| CN111680678B (en) | Target area identification method, device, equipment and readable storage medium | |
| CN113076437B (en) | Small sample image classification method and system based on label redistribution | |
| CN113378792B (en) | Weak supervision cervical cell image analysis method fusing global and local information | |
| CN112633382A (en) | Mutual-neighbor-based few-sample image classification method and system | |
| CN109145964B (en) | Method and system for realizing image color clustering | |
| Varghese et al. | INFOPLANT: Plant recognition using convolutional neural networks | |
| CN115115825B (en) | Method, device, computer equipment and storage medium for detecting object in image | |
| CN111400528A (en) | Image compression method, device, server and storage medium | |
| CN118196309B (en) | High-definition visual detection and identification system based on image processing industrial personal computer | |
| CN109492610A (en) | A kind of pedestrian recognition methods, device and readable storage medium storing program for executing again | |
| Muzakir et al. | Model for Identification and Prediction of Leaf Patterns: Preliminary Study for Improvement | |
| CN108960005B (en) | Method and system for establishing and displaying object visual label in intelligent visual Internet of things | |
| EP2383680A1 (en) | Classification of objects in harvesting applications | |
| CN113011506A (en) | Texture image classification method based on depth re-fractal spectrum network | |
| CN111144469B (en) | End-to-end multi-sequence text recognition method based on multi-dimensional associated time sequence classification neural network | |
| Anggoro et al. | Classification of Solo Batik patterns using deep learning convolutional neural networks algorithm | |
| CN115984968A (en) | Student time-space action recognition method and device, terminal equipment and medium | |
| CN111126189A (en) | Target searching method based on remote sensing image | |
| CN115019305A (en) | Method, device and equipment for identifying root tip cells and readable storage medium | |
| CN113780335A (en) | Small sample commodity image classification method, device, equipment and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200508 |