CN111291608A - Remote sensing image non-building area filtering method based on deep learning - Google Patents
Remote sensing image non-building area filtering method based on deep learning Download PDFInfo
- Publication number
- CN111291608A CN111291608A CN201911101871.1A CN201911101871A CN111291608A CN 111291608 A CN111291608 A CN 111291608A CN 201911101871 A CN201911101871 A CN 201911101871A CN 111291608 A CN111291608 A CN 111291608A
- Authority
- CN
- China
- Prior art keywords
- building
- remote sensing
- picture
- model
- deep learning
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a remote sensing image non-building area filtering method based on deep learning, which specifically comprises the following steps of marking data: marking the data of the roof type of the building, namely marking the type and the position information of different roofs of the building in the picture; marking whether the data of the building area is included, namely marking 1/0 labels on whether buildings exist in the picture or not; training a model: training a building roof segmentation model by using a Unet convolution neural network; training a building classification model by utilizing a resnet convolutional neural network; identifying a remote sensing picture with a large scale, and identifying all remote sensing images with a small scale under an area by using a building roof segmentation model when the area is identified as the area with the building; otherwise, the region is not identified; and outputting the building roof type information. The method can quickly filter out a large number of non-building areas, so that the efficiency of segmenting and identifying the roofs of urban buildings is improved, the efficiency can be improved by more than 50%, and certain resource cost is saved.
Description
Technical Field
The invention relates to the technical field of intelligent picture processing, in particular to a remote sensing image non-building area filtering method based on deep learning, which is used for quickly filtering out non-building areas in a remote sensing image.
Background
At present, when a building roof type is segmented and identified through a remote sensing image of a certain city, a small map scale, such as 3000:1, is usually selected, because the spatial resolution of the image is guaranteed to be high enough to perform accurate segmentation and identification, and then the remote sensing images of the city under the scale are identified, so that the problem exists that the city, except for a building area, also has a large number of irrelevant areas, such as water areas, vegetation areas and the like, which have no meaning and waste time for identification, and if the building area is half of the occupied city area, half of the time of model identification is invalid and can be avoided by using other technical means. Aiming at a large area of the remote sensing image which is irrelevant to a building, in order to improve the efficiency of building roof segmentation model identification, the invention has the key point that the image with a large scale is classified and identified by a convolutional neural network method so as to judge whether the area contains the building, thereby avoiding wasting a large amount of time to segment and identify the area without the building.
Disclosure of Invention
The invention provides a remote sensing image non-architectural area filtering method based on deep learning. The technical scheme of the invention is as follows:
a remote sensing image non-building area filtering method based on deep learning comprises the following steps:
100. labeling data: marking the obtained remote sensing picture;
101. training a model: training the picture recognition processing model by using the remote sensing picture marked in the step so as to enable the picture recognition processing model to obtain the subsequent capability of independently recognizing and processing the remote sensing picture;
102. inputting a large-scale remote sensing picture to be processed, and identifying and processing the trained picture identification processing model;
103. the picture identification processing model judges whether the large-scale remote sensing picture has a building or not, if the large-scale remote sensing picture does not have a building, the step 102 is returned, and if the large-scale remote sensing picture has a building, the step 104 is executed;
104. the picture identification processing model carries out building classification identification and marking on each map tile contained in the large-scale remote sensing picture;
105. and outputting the map information after the remote sensing picture is classified, identified and marked.
As a further illustration of the present invention, the picture recognition processing model comprises a building roof segmentation model and a building classification model, wherein the building classification model is used for analyzing and processing a large-scale remote sensing picture; the building roof segmentation model is used for classification identification and labeling of map tile buildings.
Further, the annotation data in step 100 includes an annotation of whether the remote sensing picture contains a building area, and an annotation of different building roof types and location information in the picture.
Further, in the step 101, a building roof segmentation model is trained by using a Unet convolutional neural network; and training a building classification model by utilizing a resnet convolutional neural network.
Further, the building roof segmentation model employs different labels for different types of building roofs in the map tiles.
Further, the types of the labels are 3 or more.
Furthermore, the scale of the large-scale remote sensing picture is a non-fixed value and is set by a user according to needs.
The present invention also provides a computer readable medium having stored thereon a computer program which is executed by a computer or a processor to implement the above-mentioned depth learning-based remote sensing image non-architectural region filtering method.
The invention has the beneficial effects that:
the method can quickly filter out a large number of non-building areas, thereby improving the efficiency of segmenting and identifying the roofs of urban buildings and saving certain resource cost. If it is assumed that the area ratio of the building area to the non-building area of a city is 1:1, the scale of the classification model processing image is 25000:1, the scale of the segmentation model processing image is 3000:1, and the recognition speed of the classification model and the segmentation model speed are 3:1, it can be estimated that about half of the time can be saved compared with the case where the segmentation recognition is directly performed on all the images of the city.
Drawings
FIG. 1 is a flow chart of the system of the present invention;
FIG. 2 is a labeled example diagram of the method labelme of the present invention.
FIG. 3 is an exemplary diagram of a large-scale remote sensing picture identified by a building classification model according to an embodiment of the present invention;
FIG. 4 is a comparison diagram of the original map tile image of the building roof segmentation model and the recognition and labeling effects in accordance with the embodiment of the present invention.
Detailed Description
Example (b):
the embodiments of the present invention will be described in detail with reference to the accompanying drawings, and it is to be understood that the described embodiments are merely a part of the embodiments of the invention, and not all of the embodiments.
In the description of the present invention, it is to be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", "first", "second", etc. indicate orientations or positional or sequential relationships based on those shown in the drawings, and are only for convenience in describing and simplifying the present invention, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
Referring to the flow chart of the attached figure 1, the flow chart of the remote sensing image non-building area filtering method based on deep learning specifically comprises the following steps:
100. labeling data: marking the data of the roof type of the building, namely marking the type and the position information of different roofs of the building in the picture; whether the data of the building area is included is marked, namely 1/0 labels are marked on whether buildings exist in the picture or not.
101. Training a model: training a building roof segmentation model by using a Unet convolution neural network; and training a building classification model by utilizing a resnet convolutional neural network.
102. After the model is trained, the remote sensing picture with a large scale can be identified, and when a building area is identified, all remote sensing images with a small scale in the area are identified by using a building roof segmentation model; otherwise the region is not identified.
103. And outputting the building roof type information.
The invention is illustrated below with specific reference to examples: firstly, a certain number of remote sensing pictures are labeled, and a building roof segmentation model and a building classification model are trained, so that the remote sensing pictures can be independently processed subsequently by the building roof segmentation model and the building classification model. The labeling in this step is manual labeling, and an existing open source labeling tool labelme can be adopted for manual labeling, referring to an example shown in fig. 2, three different types of roofs in a training remote sensing picture are labeled by using the labeling tool labelme, wherein the roof is labeled as wd, color steel tiles are labeled as wp, and a photovoltaic panel is labeled as gfb; secondly, training a building classification model by using a resnet convolutional neural network according to the marked remote sensing picture with the large scale, so that whether a building exists in the remote sensing picture with the large scale can be independently distinguished subsequently by the building classification model; and training a building roof segmentation model by using the remote sensing picture with the small scale and utilizing a Unet convolution neural network, so that the building roof segmentation model can independently identify and segment the building outline and the type in the remote sensing picture with the small scale subsequently.
After the model training is finished, other remote sensing pictures with the filtering large scale can be input to carry out formal filtering processing, whether buildings exist in the remote sensing pictures with the filtering large scale is judged and identified by the building classification model, if the buildings exist, the next step of identification and segmentation is carried out, and if the buildings do not exist, the next step of identification and segmentation is not carried out. Referring to fig. 3, the two large-scale remote sensing pictures are shown, when the building classification model identifies the left remote sensing picture, the remote sensing picture is judged to have a building and enters the next segmentation step, and when the right remote sensing picture is identified, the remote sensing picture is judged to have no link that the building does not exist and the next segmentation step is not executed.
After the building classification model identifies that the remote sensing picture has the building, the building roof segmentation model identifies and segments the remote sensing picture, specifically, the map tiles of all regions contained in the remote sensing picture with the large scale are identified and judged for the building, the building in each map tile is identified by type and outline, and the specific map tiles and the identification labeling effect are shown in the attached figure 4.
By using the method to filter the non-building area of the remote sensing picture, the time for subsequent building identification analysis processing of the remote sensing picture can be greatly reduced, and the processing speed of the remote sensing picture is effectively improved.
As can be seen from the above flow chart, the present invention also includes a computer readable medium, on which a computer program is stored, the program being executed by the data processing module to implement the above remote sensing image non-architectural region filtering method based on deep learning.
The foregoing is illustrative of the preferred embodiments of the present invention only and is not to be construed as limiting the claims. The invention is not limited to the above embodiments, the specific construction of which allows variations, and in any case variations, which are within the scope of the invention as defined in the independent claims.
Claims (8)
1. A remote sensing image non-building area filtering method based on deep learning is characterized in that: the method comprises the following steps:
100. labeling data: marking the obtained remote sensing picture;
101. training a model: training the picture recognition processing model by using the remote sensing picture marked in the step so as to enable the picture recognition processing model to obtain the subsequent capability of independently recognizing and processing the remote sensing picture;
102. inputting a large-scale remote sensing picture to be processed, and identifying and processing the trained picture identification processing model;
103. the picture identification processing model judges whether the large-scale remote sensing picture has a building or not, if the large-scale remote sensing picture does not have a building, the step 102 is returned, and if the large-scale remote sensing picture has a building, the step 104 is executed;
104. the picture identification processing model carries out building classification identification and marking on each map tile contained in the large-scale remote sensing picture;
105. and outputting the map information after the remote sensing picture is classified, identified and marked.
2. The remote sensing image non-architectural region filtering method based on deep learning of claim 1, wherein: the picture identification processing model comprises a building roof segmentation model and a building classification model, and the building classification model is used for analyzing and processing a large-scale remote sensing picture; the building roof segmentation model is used for classification identification and labeling of map tile buildings.
3. The remote sensing image non-architectural region filtering method based on deep learning of claim 1, wherein: the annotation data in step 100 includes an annotation to determine whether the remote sensing picture contains a building area, and an annotation to identify different types and locations of the roof of the building in the picture.
4. The remote sensing image non-architectural region filtering method based on deep learning of claim 1, wherein: in the step 101, a building roof segmentation model is trained by using a Unet convolutional neural network; and training a building classification model by utilizing a resnet convolutional neural network.
5. The remote sensing image non-architectural region filtering method based on deep learning of claim 1, wherein: the building roof segmentation model employs different labels for different types of building roofs in the map tiles.
6. The remote sensing image non-architectural region filtering method based on deep learning of claim 5, wherein: the types of the labels are 3 or more than three.
7. The remote sensing image non-architectural region filtering method based on deep learning of claim 1, wherein: the scale of the large-scale remote sensing picture is a non-fixed value and is set by a user according to needs.
8. A computer-readable medium having a computer program stored thereon, characterized in that: the program being executable by a computer or processor to implement the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911101871.1A CN111291608A (en) | 2019-11-12 | 2019-11-12 | Remote sensing image non-building area filtering method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911101871.1A CN111291608A (en) | 2019-11-12 | 2019-11-12 | Remote sensing image non-building area filtering method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111291608A true CN111291608A (en) | 2020-06-16 |
Family
ID=71021534
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911101871.1A Pending CN111291608A (en) | 2019-11-12 | 2019-11-12 | Remote sensing image non-building area filtering method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111291608A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200029A (en) * | 2020-09-27 | 2021-01-08 | 电子科技大学 | Remote sensing image building extraction method based on improved UNet + + network |
| CN112580484A (en) * | 2020-12-14 | 2021-03-30 | 中国农业大学 | Corn straw coverage identification method and device based on deep learning remote sensing image |
| CN115457281A (en) * | 2022-11-10 | 2022-12-09 | 珠高智能科技(深圳)有限公司 | Roof image set generation method and device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090310867A1 (en) * | 2008-06-12 | 2009-12-17 | Bogdan Calin Mihai Matei | Building segmentation for densely built urban regions using aerial lidar data |
| US20170076438A1 (en) * | 2015-08-31 | 2017-03-16 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| CN109583425A (en) * | 2018-12-21 | 2019-04-05 | 西安电子科技大学 | A kind of integrated recognition methods of the remote sensing images ship based on deep learning |
-
2019
- 2019-11-12 CN CN201911101871.1A patent/CN111291608A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090310867A1 (en) * | 2008-06-12 | 2009-12-17 | Bogdan Calin Mihai Matei | Building segmentation for densely built urban regions using aerial lidar data |
| US20170076438A1 (en) * | 2015-08-31 | 2017-03-16 | Cape Analytics, Inc. | Systems and methods for analyzing remote sensing imagery |
| CN109583425A (en) * | 2018-12-21 | 2019-04-05 | 西安电子科技大学 | A kind of integrated recognition methods of the remote sensing images ship based on deep learning |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200029A (en) * | 2020-09-27 | 2021-01-08 | 电子科技大学 | Remote sensing image building extraction method based on improved UNet + + network |
| CN112200029B (en) * | 2020-09-27 | 2022-03-25 | 电子科技大学 | Remote sensing image building extraction method based on improved UNet + + network |
| CN112580484A (en) * | 2020-12-14 | 2021-03-30 | 中国农业大学 | Corn straw coverage identification method and device based on deep learning remote sensing image |
| CN112580484B (en) * | 2020-12-14 | 2024-03-29 | 中国农业大学 | Remote sensing image corn straw coverage recognition method and device based on deep learning |
| CN115457281A (en) * | 2022-11-10 | 2022-12-09 | 珠高智能科技(深圳)有限公司 | Roof image set generation method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3806064B1 (en) | Method and apparatus for detecting parking space usage condition, electronic device, and storage medium | |
| CN111291608A (en) | Remote sensing image non-building area filtering method based on deep learning | |
| CN110348441B (en) | Value-added tax invoice identification method and device, computer equipment and storage medium | |
| CN109522900B (en) | Natural scene character recognition method and device | |
| CN112101357A (en) | RPA robot intelligent element positioning and picking method and system | |
| CN107480603B (en) | Synchronous mapping and object segmentation method based on SLAM and depth camera | |
| CN112508975A (en) | Image identification method, device, equipment and storage medium | |
| CN108062529B (en) | Intelligent identification method for chemical structural formula | |
| CN111191500A (en) | Photovoltaic roof resource identification method based on deep learning image segmentation | |
| CN111399638B (en) | Blind computer and intelligent mobile phone auxiliary control method suitable for blind computer | |
| CN102693334B (en) | Based on the dynamic component recognition methods of CAD electronic drawing | |
| CN110675940A (en) | Pathological image labeling method and device, computer equipment and storage medium | |
| CN109359043B (en) | Mobile game automatic testing method based on machine learning | |
| US20230129263A1 (en) | Multi-level transferable region-based domain adaptive object detection apparatus and method | |
| CN111028261B (en) | High-precision semi-automatic image data annotation method, electronic device and storage medium | |
| CN111008576A (en) | Pedestrian detection and model training and updating method, device and readable storage medium thereof | |
| CN112633118A (en) | Text information extraction method, equipment and storage medium | |
| CN110910414B (en) | Image contour generation method, image labeling method, electronic device and storage medium | |
| JP2009217832A (en) | Method and device for automatically recognizing road sign in video image, and storage medium which stores program of road sign automatic recognition | |
| CN114067339A (en) | Image recognition method and device, electronic equipment and computer readable storage medium | |
| CN103295238A (en) | ROI (region of interest) motion detection based real-time video positioning method for Android platform | |
| CN115546824B (en) | Taboo picture identification method, apparatus and storage medium | |
| CN112241749A (en) | Character recognition model training method, device and equipment | |
| KR101411893B1 (en) | Automatic Recognition Method of Direction Information in Road Sign Image | |
| CN116543327A (en) | Method, device, computer equipment and storage medium for identifying work types of operators |
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 |