CN113033385A - Deep learning-based violation building remote sensing identification method and system - Google Patents
Deep learning-based violation building remote sensing identification method and system Download PDFInfo
- Publication number
- CN113033385A CN113033385A CN202110309208.1A CN202110309208A CN113033385A CN 113033385 A CN113033385 A CN 113033385A CN 202110309208 A CN202110309208 A CN 202110309208A CN 113033385 A CN113033385 A CN 113033385A
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- building
- deep learning
- image
- illegal
- 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
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for identifying violation buildings based on deep learning, which are characterized by comprising the steps of carrying out convolution, pooling and deconvolution on vector data of a remote sensing image, extracting house characteristics to obtain a vector data characteristic diagram, carrying out vectorization on house outline pixels, converting the vector data characteristic diagram into a grid image and using the grid image for model training; inputting the grid image into a deep learning network to obtain a building contour training model; the remote sensing image to be tested is input into the deep learning network as a test set, the remote sensing image classification output result is carried out on the test set, the target identification of the remote sensing image is realized, a new detection and monitoring technical means is provided for the rapid and accurate identification of the illegal buildings through the combination of remote sensing image analysis, deep learning and visual big data identification technologies, the urban management department can be helped to improve the identification efficiency of the illegal buildings, the illegal buildings can be found in time, the management cost is reduced, and the urban management level is improved.
Description
Technical Field
The invention relates to the technical field of automatic detection of urban illegal buildings, in particular to a remote sensing identification method and system of illegal buildings based on deep learning.
Background
At present, urban illegal building detection mainly depends on manual comparison and investigation of remote sensing images with 0.3 m resolution, and numerous scholars provide a plurality of schemes for solving the problems of limited human resources and long time consumption of manual comparison. The Zhanjin Rui proposes a violation building detection overall architecture which is based on the satellite images of the past year and combines the field mapping and planning information; the Jujiawei and the like provide a system for constructing a three-dimensional model by using an unmanned aerial vehicle low-altitude environment-shot image so as to identify a violation building. In the field of remote sensing image change detection, Malila firstly proposes a Change Vector Analysis (CVA), and describes each pixel as a one-dimensional column vector by using multiband remote sensing image data, so as to calculate a vector difference value between pixels at the same position in the front period and the rear period, and represent the change intensity; and the yellow dimension and the like are combined with Principal Component Analysis (PCA) and Change Vector Analysis (CVA) to carry out difference value operation and threshold division after extracting the first principal component from the multiband image data, so that the influence of image noise is reduced. In recent years, with the improvement of the computing performance of a computer, a monitoring classification method based on deep learning often obtains better effect in practice, and by Zhang Xiaodong and the like, a mainstream network fast R-CNN in a target detection field is applied to high-resolution remote sensing image change detection to obtain ideal effect. Scholars at home and abroad contribute effective system architecture and technical methods for detecting the illegal buildings, however, the traditional change detection algorithm based on pixels is easy to generate the phenomenon of 'salt and pepper' and is difficult to meet the precision requirement of rough screening of the illegal buildings, and the operation time based on deep learning highly depends on the performance of a computer and is difficult to be applied to a real-time detection system, so how to effectively and quickly realize the automation of the illegal building identification work is still to be continuously researched. The problem that the identification efficiency is not high in the prior art.
Disclosure of Invention
Aiming at the problems in the background technology, a brand-new violation building remote sensing identification method based on deep learning is provided. By utilizing a strong and effective characteristic extraction technology of a deep network, the problem that the existing picture identification technology cannot meet the identification of hidden danger of illegal buildings caused by environmental changes, such as self-built buildings of houses, self-built buildings of roofs, self-built buildings of courtyards, self-built buildings of balconies, self-built buildings of public channels and the like, can be at least partially solved through picture similarity analysis.
The invention relates to a violation building remote sensing identification method based on deep learning, which comprises the following steps:
s1, obtaining remote sensing images of buildings and roads in a ground environment, performing convolution, pooling and deconvolution on vector data of the remote sensing images, extracting house features to obtain a vector data feature map, vectorizing house contour pixels, and converting the vector data feature map into a grid image for model training;
s2, inputting the raster image into a deep learning network to obtain a building contour training model;
s3, inputting the remote sensing image to be tested into the deep learning network as a test set, and performing remote sensing image classification output on the test set to realize target identification of the remote sensing image.
The invention provides a new detection and monitoring technical means for the rapid and accurate identification of the illegal buildings by combining the remote sensing image analysis, the deep learning and the visual big data identification technology, and can help the city management department to improve the identification efficiency of the illegal buildings, discover the illegal buildings in time, reduce the management cost and improve the city management level.
Specifically, the step of performing remote sensing image classification on the test set to output the result includes: and comparing the difference image and the similarity of the original data set and the test set by a structural similarity algorithm, judging whether the illegal building appears according to the maximum value of the image similarity, and prompting the position of the illegal building.
Further, carrying out mean value fuzzy processing on the original data set and the test set, and converting the mean value fuzzy processing into a gray level image; denoising the gray level image;
and comparing the difference graph by a structural similarity algorithm to obtain a difference graph, detecting the outline of the difference graph and marking the similarity, and alarming when the similarity is lower than a set threshold value.
Further, the building contour training model can be optimized by the following steps: acquiring violation building images at a plurality of angles and a plurality of distances in the remote sensing image as training data;
generating a clear illegal building data sample, and labeling the building to obtain a labeled data set;
training the labeled data set to obtain a training model,
and calculating the remote sensing image to be detected through a training model to obtain the multiple violation building types and coordinates in the image.
Further, the illegal building marking comprises: the building is built by oneself to the house, the building is built by oneself to the roof, the building is built by oneself in courtyard, the building is built by oneself to the balcony, the building is built by oneself to the public channel.
The deep learning network, namely the convolution neural network, is a feedforward neural network, and the artificial neuron can respond to peripheral units and can perform large-scale image processing. The convolutional neural network includes convolutional layers and pooling layers.
Further, the building contour training model is packaged as a building contour extraction tool.
Further, the present invention provides a readable storage medium having a control program stored thereon, characterized in that: when being executed by a processor, the control program realizes the remote sensing identification method for the illegal buildings based on deep learning.
Further, the present invention provides a computer control system, including a storage, a processor, and a control program stored in the storage and executable by the processor, wherein: the processor executes the control program to realize the remote sensing identification method for the illegal buildings based on deep learning.
In order that the invention may be more clearly understood, specific embodiments thereof will be described hereinafter with reference to the accompanying drawings.
Drawings
Fig. 1 is a flow chart of a violation building remote sensing identification method based on deep learning in the embodiment of the invention.
Detailed Description
Referring to fig. 1, a flow chart of a remote sensing identification method for illegal buildings based on deep learning in the embodiment of the invention is shown.
S1, obtaining remote sensing images of buildings and roads in a ground environment, performing convolution, pooling and deconvolution on vector data of the remote sensing images, extracting house features to obtain a vector data feature map, vectorizing house contour pixels, and converting the vector data feature map into a grid image for model training;
s2, inputting the raster image into a deep learning network to obtain a building contour training model;
s3, inputting the remote sensing image to be tested into the deep learning network as a test set, and performing remote sensing image classification output on the test set to realize target identification of the remote sensing image.
The invention provides a new detection and monitoring technical means for the rapid and accurate identification of the illegal buildings by combining the remote sensing image analysis, the deep learning and the visual big data identification technology, and can help the city management department to improve the identification efficiency of the illegal buildings, discover the illegal buildings in time, reduce the management cost and improve the city management level.
Specifically, the step of performing remote sensing image classification on the test set to output the result includes: and comparing the difference image and the similarity of the original data set and the test set by a structural similarity algorithm, judging whether the illegal building appears according to the maximum value of the image similarity, and prompting the position of the illegal building.
Further, carrying out mean value fuzzy processing on the original data set and the test set, and converting the mean value fuzzy processing into a gray level image; denoising the gray level image;
the target pixel is given a matrix on the image, the matrix comprises 8 surrounding pixels taking the target pixel as the center, the matrix forms a convolution kernel, the average value of all pixels in the matrix replaces the value of the target pixel, and the process of blurring the average value is the process of moving the matrix from left to right and from top to bottom, and one pixel is moved each time. The image blurring processing is to process the image more blurry, blur part of details irrelevant to the image, process side effects caused by noise, filter objects with small size and brightness, smooth the image and facilitate target extraction. In nature, RGB colors are easily affected by illumination, and gradient information can provide more essential features. The image after the graying processing only contains brightness information and does not contain color information, so that the picture is convenient to store. Three channels are changed into one channel, so that the processing efficiency is improved. The on operation processing of the gray level image can remove noise, eliminate the detail part smaller than the structural element in the image and keep the local shape of the object unchanged. When the object is brighter than the background, the small-area object can be eliminated, isolated points higher than the adjacent points are eliminated, the object contour is smoothed, and a narrower narrow neck is broken.
And comparing the difference graph by a structural similarity algorithm to obtain a difference graph, detecting the outline of the difference graph and marking the similarity, and alarming when the similarity is lower than a set threshold value.
The preprocessing ensures that the black and white of the image are clear, the contrast is increased, and simultaneously the problems of spots, pocks and the like caused by sharpening processing are avoided, so that the image not only keeps smooth natural color tone, but also can emphasize the contrast of the changed details. In order to improve the precision, the influences of light, plant colors and the like are ignored, small differences are removed, large differences are reserved, and when the large differences exist, the fact that the picture to be detected and the reference picture have abnormal changes is proved, and the abnormality can be a hidden danger of illegal construction.
Further, the building contour training model can be optimized by the following steps: acquiring violation building images at a plurality of angles and a plurality of distances in the remote sensing image as training data;
generating a clear illegal building data sample, and labeling the building to obtain a labeled data set;
training the labeled data set to obtain a training model,
and calculating the remote sensing image to be detected through a training model to obtain the multiple violation building types and coordinates in the image.
Further, the illegal building marking comprises: the building is built by oneself to the house, the building is built by oneself to the roof, the building is built by oneself in courtyard, the building is built by oneself to the balcony, the building is built by oneself to the public channel.
The deep learning network, namely the convolution neural network, is a feedforward neural network, and the artificial neuron can respond to peripheral units and can perform large-scale image processing. The convolutional neural network includes convolutional layers and pooling layers.
Further, the building contour training model is packaged as a building contour extraction tool.
According to the method, the real-time snapshot picture is compared with the reference picture, if the similarity is lower than the set limit value, the difference part is found out, the picture marking is carried out, and the early warning of the hidden danger of the illegal building is realized, wherein the accuracy of the early warning can be further improved by adjusting the limit value of the difference according to different environments. The image recognition needs a large number of training samples to carry out deep learning, the recognition accuracy can be improved, otherwise, real hidden dangers cannot be recognized, the image similarity comparison is compared with normal images, the normal images are easy to obtain in the system operation process, the longer the system operation time is, the more the normal images are, and the more accurate the result is compared through the similarity. Once a real hidden trouble occurs, the similarity between the photographed picture and the reference picture is greatly reduced. The invention is based on a deep learning method to realize hidden danger early warning of illegal buildings. Simulation experiments show that the method can better remove the interference background in the image, only images a specific target object, has high precision, and is feasible by verifying the remote sensing identification method of the illegal buildings based on deep learning.
Compared with the traditional imaging method, the method provided by the invention provides a new detection and monitoring technical means for the rapid and accurate identification of the illegal building through the combination of remote sensing image analysis, deep learning and visual big data identification technology, and can help the city management department to improve the identification efficiency of the illegal building, discover the illegal building in time, reduce the management cost and improve the city management level.
The present invention is not limited to the above-described embodiments, and various modifications and variations of the present invention are included in the scope of the claims and the equivalent technology of the present invention if they do not depart from the spirit and scope of the present invention.
Claims (8)
1. A violation building remote sensing identification method based on deep learning comprises the following steps:
obtaining remote sensing images of buildings and roads in a ground environment as an original data set, performing convolution, pooling and deconvolution on vector data of the remote sensing images, extracting house features to obtain a vector data feature map, vectorizing house contour pixels, and converting the vector data feature map into a grid image for model training;
inputting the raster image into a deep learning network to obtain a building contour training model;
and inputting the remote sensing image to be tested into the deep learning network as a test set, and performing remote sensing image classification output on the test set to realize target identification of the remote sensing image.
2. The remote sensing illegal building identification method based on deep learning of claim 1, wherein the step of performing remote sensing image classification on the test set and outputting the result comprises: and comparing the difference image and the similarity of the original data set and the test set by a structural similarity algorithm, judging whether the illegal building appears according to the maximum value of the image similarity, and prompting the position of the illegal building.
3. The remote sensing identification method for the illegal building based on the deep learning of the claim 1 is characterized in that mean fuzzy processing is carried out on the original data set and the test set, and the mean fuzzy processing is converted into a gray scale image; denoising the gray level image;
and comparing the difference graph by a structural similarity algorithm to obtain a difference graph, detecting the outline of the difference graph and marking the similarity, and alarming when the similarity is lower than a set threshold value.
4. The remote sensing identification method for the illegal building based on the deep learning of the claim 1 is characterized in that: the building outline training model can be optimized through the following steps: acquiring violation building images at a plurality of angles and a plurality of distances in the remote sensing image as training data;
generating a clear illegal building data sample, and labeling the building to obtain a labeled data set;
training the labeled data set to obtain a training model,
and calculating the remote sensing image to be detected through a training model to obtain the multiple violation building types and coordinates in the image.
5. The remote sensing identification method for the illegal building based on the deep learning of the claim 1 is characterized in that: the illegal building marking comprises the following steps: the building is built by oneself to the house, the building is built by oneself to the roof, the building is built by oneself in courtyard, the building is built by oneself to the balcony, the building is built by oneself to the public channel.
6. The remote sensing identification method for the illegal building based on the deep learning of the claim 1 is characterized in that: and packaging the building outline training model into a building outline extraction tool.
7. A readable storage medium having a control program stored thereon, characterized in that: the control program is executed by a processor to realize the remote sensing identification method of the illegal buildings based on deep learning of any one of claims 1 to 6.
8. A computer control system comprising a memory, a processor, and a control program stored in said memory and executable by said processor, characterized in that: the processor executes the control program to realize the remote sensing identification method for the illegal buildings based on deep learning according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110309208.1A CN113033385A (en) | 2021-03-23 | 2021-03-23 | Deep learning-based violation building remote sensing identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110309208.1A CN113033385A (en) | 2021-03-23 | 2021-03-23 | Deep learning-based violation building remote sensing identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113033385A true CN113033385A (en) | 2021-06-25 |
Family
ID=76472907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110309208.1A Pending CN113033385A (en) | 2021-03-23 | 2021-03-23 | Deep learning-based violation building remote sensing identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113033385A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113989680A (en) * | 2021-12-27 | 2022-01-28 | 苏州工业园区测绘地理信息有限公司 | Automatic building three-dimensional scene construction method and system |
| CN114120120A (en) * | 2021-11-25 | 2022-03-01 | 广东电网有限责任公司 | Method, device, equipment and medium for detecting illegal building based on remote sensing image |
| CN115631362A (en) * | 2022-09-26 | 2023-01-20 | 北京霍里思特科技有限公司 | Ironware identification method and ironware identification device |
| CN115641415A (en) * | 2022-12-26 | 2023-01-24 | 成都国星宇航科技股份有限公司 | Method, device, equipment and medium for generating three-dimensional scene based on satellite image |
-
2021
- 2021-03-23 CN CN202110309208.1A patent/CN113033385A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114120120A (en) * | 2021-11-25 | 2022-03-01 | 广东电网有限责任公司 | Method, device, equipment and medium for detecting illegal building based on remote sensing image |
| CN113989680A (en) * | 2021-12-27 | 2022-01-28 | 苏州工业园区测绘地理信息有限公司 | Automatic building three-dimensional scene construction method and system |
| CN113989680B (en) * | 2021-12-27 | 2022-03-29 | 苏州工业园区测绘地理信息有限公司 | Automatic building three-dimensional scene construction method and system |
| CN115631362A (en) * | 2022-09-26 | 2023-01-20 | 北京霍里思特科技有限公司 | Ironware identification method and ironware identification device |
| CN115641415A (en) * | 2022-12-26 | 2023-01-24 | 成都国星宇航科技股份有限公司 | Method, device, equipment and medium for generating three-dimensional scene based on satellite image |
| CN115641415B (en) * | 2022-12-26 | 2023-08-22 | 成都国星宇航科技股份有限公司 | Method, device, equipment and medium for generating three-dimensional scene based on satellite image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sirmacek et al. | Urban-area and building detection using SIFT keypoints and graph theory | |
| CN113033385A (en) | Deep learning-based violation building remote sensing identification method and system | |
| CN105404886B (en) | Characteristic model generation method and characteristic model generating means | |
| Lei et al. | Automated pavement distress detection and deterioration analysis using street view map | |
| CN110033431B (en) | Non-contact detection device and detection method for detecting corrosion area on surface of steel bridge | |
| Hassaan et al. | Precision forestry: trees counting in urban areas using visible imagery based on an unmanned aerial vehicle | |
| CN109410171B (en) | Target significance detection method for rainy image | |
| CN113160150B (en) | AI (Artificial intelligence) detection method and device for invasion of foreign matters in wire mesh | |
| US20130342694A1 (en) | Method and system for use of intrinsic images in an automotive driver-vehicle-assistance device | |
| CN114898097B (en) | Image recognition method and system | |
| Chen et al. | Object-based multi-modal convolution neural networks for building extraction using panchromatic and multispectral imagery | |
| CN111274964B (en) | Detection method for analyzing water surface pollutants based on visual saliency of unmanned aerial vehicle | |
| Palenichka et al. | Multi-scale segmentation of forest areas and tree detection in LiDAR images by the attentive vision method | |
| CN112907626A (en) | Moving object extraction method based on satellite time-exceeding phase data multi-source information | |
| CN116433940A (en) | Remote sensing image change detection method based on twin mirror network | |
| CN116258956A (en) | Unmanned aerial vehicle tree recognition method, unmanned aerial vehicle tree recognition equipment, storage medium and unmanned aerial vehicle tree recognition device | |
| CN113033386B (en) | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system | |
| Yadav et al. | Supervised learning based greenery region detection using unnamed aerial vehicle for smart city application | |
| CN113378642A (en) | Method for detecting illegal occupation buildings in rural areas | |
| Heidary et al. | Urban change detection by fully convolutional siamese concatenate network with attention | |
| CN115019306A (en) | Embedding box label batch identification method and system based on deep learning and machine vision | |
| CN112036246B (en) | Construction method of remote sensing image classification model, remote sensing image classification method and system | |
| Altaei et al. | Satellite image classification using multi features based descriptors | |
| CN113963270A (en) | High resolution remote sensing image building detection method | |
| Golej et al. | Detection of people and vehicles using very high-resolution satellite images |
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 |