CN111598048A - Urban village-in-village identification method integrating high-resolution remote sensing image and street view image - Google Patents
Urban village-in-village identification method integrating high-resolution remote sensing image and street view image Download PDFInfo
- Publication number
- CN111598048A CN111598048A CN202010481052.0A CN202010481052A CN111598048A CN 111598048 A CN111598048 A CN 111598048A CN 202010481052 A CN202010481052 A CN 202010481052A CN 111598048 A CN111598048 A CN 111598048A
- Authority
- CN
- China
- Prior art keywords
- street
- remote sensing
- resolution remote
- image
- street view
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- 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)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the field of geographic images, and provides a city-village identification method fusing a high-resolution remote sensing image and a street view image, which comprises the following steps: obtaining a high-resolution remote sensing image of a selected area for pre-extraction and multi-scale segmentation; acquiring multi-view street view images of the selected area, and constructing an optimal street space quality evaluation model and a corresponding optimal street view image feature combination mode under each view; extracting the characteristics of the multi-class high-resolution remote sensing images of the parcels to be classified; extracting a plurality of street quality features of each parcel to be classified; and performing heterogeneous fusion on the multi-class high-resolution remote sensing image features and the street quality features, and judging whether the rest plots are rural-in-urban areas or not by using a machine learning algorithm. The method integrates the street view image information into the urban village extraction process based on the high-resolution remote sensing image, constructs a feature space with higher discrimination, and improves the urban village identification precision.
Description
Technical Field
The invention relates to the technical field of geographic images, in particular to a method for identifying villages in cities by fusing high-resolution remote sensing images and street view images.
Background
Urban Village (UV) is an irregular living space in a city, is free from a city management system, and is a special product in the process of urbanization in china. Compared with the conventional urban built-up area, the urban rural internal street space has low quality, has the characteristics of disordered land utilization conditions, poor building density and high quality, lack of infrastructure, dirty environment and the like, and has negative effects on urban landscape and public health. With the advance of novel urbanization construction in China, high-quality development becomes the subject of a new period. The method has the advantages that the spatial distribution of urban villages and the environmental quality information thereof are timely and accurately acquired, and the method has important significance for optimizing urban space and improving human living environment.
High-resolution remote sensing images (high-resolution remote sensing images) have become important data sources for city planning and management due to the advantages of wide observation range, abundant ground feature information, convenient acquisition and the like, and high-resolution remote sensing images and related technologies are an effective means for acquiring urban-rural distribution information. The high-resolution remote sensing image is a reliable data source for efficiently acquiring the spatial distribution of the urban village at present, however, due to the complexity and the landscape diversity of the urban village, the single high-resolution remote sensing image cannot meet the high-precision requirement of urban village identification.
Disclosure of Invention
The invention aims to provide a method for identifying villages in cities by fusing high-resolution remote sensing images and street view images, which aims to solve the problems in the prior art.
In order to achieve the purpose, the technical scheme adopted by the invention comprises the following steps:
the method comprises the following steps: acquiring a high-resolution remote sensing image of a selected area for preprocessing, extracting a region to be classified from the preprocessed high-resolution remote sensing image according to the land utilization condition of the selected area, and performing multi-scale segmentation on the region to be classified to obtain a plurality of plots to be classified;
step two: acquiring and labeling multi-view street view images of the selected area, extracting a plurality of street view image features, constructing a street space quality evaluation model on a training set by utilizing a machine learning algorithm, and selecting an optimal street space quality evaluation model and a corresponding optimal street view image feature combination mode under each view;
step three: extracting the characteristics of various high-resolution remote sensing images of the plots to be classified from the preprocessed high-resolution remote sensing images according to the characteristics of the plots in the rural areas in the city;
step four: drawing the street space quality of the selected area by using the optimal street space quality evaluation model under each visual angle and a corresponding characteristic combination mode, and acquiring a spatially continuous street space quality distribution map of the selected area by using a spatial interpolation strategy so as to extract a plurality of street space quality indexes in the plots to be classified;
step five: and performing heterogeneous fusion on the multi-class high-resolution remote sensing image features and the street quality features of each plot to be classified, constructing a classifier by using a machine learning algorithm, and judging whether the rest plots are villages in the city or not.
Optionally, in the first step, the high-resolution remote sensing image is from a GF-2PMS1 sensor and includes a panchromatic band and a multispectral band, and the preprocessing step includes: orthorectification, radiometric calibration, atmospheric rectification, image registration and image fusion.
Optionally, in the first step, the land utilization conditions of the selected area include water, vegetation, roads, urban village building areas and non-urban village building areas, wherein the areas to be classified are the urban village areas and the non-urban village building areas.
Optionally, in the first step, performing multi-scale segmentation on the region to be classified includes: and constructing a partition layer with two scales of a building object level and a land block level by taking the road data and the vegetation data as constraint conditions, and performing multi-scale partition on the to-be-classified area.
Optionally, in the second step, the plurality of streetscape image features include accelerated robust features, histogram of oriented gradients features and semantic features, and the machine learning algorithm includes a support vector machine and a random forest algorithm.
Optionally, in the second step, a machine learning algorithm is used to construct a street space quality evaluation model on the training set, and selecting an optimal street space quality evaluation model and a corresponding optimal street view image feature combination mode at each view angle includes: dividing the street view image features into a training set and a test set, constructing a plurality of street space quality evaluation models on the training set by using the machine learning algorithm based on the training set, measuring model performance by using the classification precision of each model on the test set and the Kappa coefficient, and selecting the optimal street space quality model and the corresponding optimal street view image feature combination mode under each view angle.
Optionally, in the third step, the categories of the multiple types of high-resolution remote sensing image features include spectral features, shape features, texture features, building structures and scene features, and each category corresponds to multiple high-resolution remote sensing image features.
Optionally, in the fourth step, the street quality features are a mean, a variance, a maximum, a minimum and a range of street space quality within each parcel to be classified.
Optionally, in the fifth step, the machine learning algorithm is a random forest algorithm.
Optionally, after the fifth step, the identification method further includes a sixth step of: and comprehensively acquiring the urban-rural spatial distribution condition of the plot scale by combining field investigation, a Google Earth map and a visual interpretation technology, taking the urban-rural spatial distribution condition as a ground true value, and evaluating the model performance by utilizing the integral classification precision, the Kappa coefficient, the user precision and the producer precision of the urban-rural plot.
The invention has the beneficial effects that: the method is characterized in that street view image information is integrated into a high-resolution remote sensing image-based urban village extraction process, so that the spectrum, shape, texture, building structure and scene characteristics provided by the aerial view high-resolution remote sensing image are comprehensively complementary with street space quality information provided by the street view image of the human visual angle, a characteristic space with higher discrimination is constructed, and the urban village identification precision is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed for the embodiments or the prior art descriptions 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 diagram illustrating GF-2 true color images and pre-classification results according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a high/low spatial quality street view image at different viewing angles according to an embodiment of the present invention.
Fig. 3 is a schematic view of a process of identifying villages in a city according to an embodiment of the present invention.
Fig. 4 is a comparison graph of the comprehensive urban village extraction results based on different images according to the embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantageous effects to be solved by the present invention more clearly apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In view of the fact that a single high-resolution remote sensing image is difficult to meet the high-precision requirement of urban village identification, the method adopts a multi-source data fusion technology to improve the urban village identification precision. The main difference between the urban village and the conventional urban built-up area is that the built-up environment inside the urban village, namely the street space, is low in quality. However, the built environment inside the urban village is difficult to directly reflect in the remote sensing image based on the overlooking visual angle. The street view map provides 360-degree panoramic images of streets and has the advantages of wide coverage range, high position precision, low acquisition cost and the like. The basic idea of the invention is to perform urban village identification by fusing the high-resolution remote sensing image and the street view image.
In this embodiment, a overseas district in Guangzhou city is selected as a research area, and the method for identifying villages in cities for identifying the districts in cities in the district, which is provided by the invention and combines the high-resolution remote sensing image and the streetscape image, specifically comprises the following steps.
The method comprises the following steps: the method comprises the steps of obtaining high-resolution remote sensing images of a selected area for preprocessing, extracting areas to be classified from the preprocessed high-resolution remote sensing images according to the land utilization condition of the selected area, and conducting multi-scale segmentation on the areas to be classified to obtain a plurality of plots to be classified.
Guangzhou city is located in the central and south of Guangdong province and is one of the most important cities forming the city village in China at first. Guangzhou city belongs to marine subtropical monsoon climate, and is evergreen in all seasons, and the landscape such as urban interior vegetation is relatively stable. 11 districts are managed in Guangzhou city, and the Virginia area (113.24 degrees E-113.32 degrees E, 23.11 degrees N-23.17 degrees N) is the urban area with the smallest area and the highest population density in the 11 urban areas in Guangzhou city, and comprises typical urban villages such as dengfeng, west pits, Yaotai and the like.
The high-resolution remote sensing image used in the embodiment is from a GF-2PMS1 sensor and comprises a panchromatic band with 1m spatial resolution and a multispectral band (blue, green, red and near infrared) with 4m spatial resolution, and the acquisition time of the image is 2017, 9 and 15 days.
The GF-2 image preprocessing comprises the following steps: orthorectification, radiometric calibration, atmospheric rectification, image registration and image fusion. Firstly, carrying out orthorectification by utilizing self-contained parameters of full-color and multispectral images, secondly, carrying out radiometric calibration on multispectral data by adopting a GF-2 absolute radiometric calibration coefficient, and carrying out atmospheric correction on a radiance image after multispectral radiometric calibration by utilizing a FLAASH module in ENVI 5.3. And performing image registration by taking the panchromatic image as a reference, selecting a ground control point from the image to be corrected for geometric correction, and finally performing image fusion on corrected multispectral data and panchromatic data by using a Gram-Schmidt method, wherein the spatial resolution of the fused image is 1 m.
The land utilization conditions of the overseas region include water, vegetation, roads, urban village building regions and non-urban village building regions. The Vegetation distribution condition of the overseas region is obtained by adopting a Normalized Difference Vegetation Index (NDVI), and the NDVI is more than or equal to 0.36 and is used as a threshold value to extract the Vegetation region. Road network data and water vector data of the overseas region come from OpenStreetMap (OSM), and 4m buffer regions are generated on two sides of a linear road network according to the width of a conventional lane to serve as road network planar data. The pre-extraction result of the land cover type of the overseas region is shown in figure 1, and then only the rural areas and non-rural building areas in the region to be classified need to be distinguished.
The present embodiment employs an Object-Based Image Analysis (OBIA) method to extract villages in town. In view of the complexity of the rural area in the city, the high-resolution remote sensing image is segmented in a multi-scale mode to simultaneously take macroscopic and microscopic features into consideration. Relatively wide roads are usually arranged between the rural areas in the city and other types of places around the rural areas, so that the segmentation accuracy can be improved by utilizing a road network provided by the OSM to assist the image segmentation process on the basis of the traditional multi-scale segmentation method. Considering that the Guangzhou city street trees are dense and can also be used as partitions of villages in cities and other areas, a continuously distributed vegetation data constraint segmentation process is added. Three parameters of dimension, shape and compactness are required to be set in the multi-dimension segmentation process. The embodiment constructs two-scale segmentation layers of a building object level and a plot level for extracting the village in the city. And selecting proper segmentation parameters by using a trial and error method, wherein the segmentation parameters of the final land block level are 320/0.9/0.8, and the segmentation parameters of the building object level are 100/0.8/0.5. The GF-2 image multi-scale segmentation process is completed based on an eCongnition development 9.0 platform, and feature extraction and urban village identification are carried out by taking the segmentation result of the land parcel level as a statistical unit.
Step two: and acquiring and labeling multi-view street view images of the selected area, extracting a plurality of street view image features, constructing a street space quality evaluation model on a training set by utilizing a machine learning algorithm, and selecting an optimal street view space quality evaluation model and a corresponding optimal street view image feature combination mode under each view.
The street view image used in the embodiment is derived from a hundred-degree map, sampling points are generated on the offsite road network at intervals of 50m, 4 street view images which are parallel to the road (front and back) and perpendicular to the road direction (left and right) are respectively obtained according to the orientation of the road where the sampling points are located, the field angle of each image is set to 90 degrees, and the street view images can fully cover the surrounding environment of the sampling points. And finally, 59720 street view images on 14930 sampling points of the overseall district in Guangzhou city are obtained, wherein the actual shooting time of the street view images is 2017, 5 months and is basically consistent with the acquisition time of the high-resolution remote sensing images.
The low street space quality is a main characteristic of distinguishing urban villages from conventional urban built-up areas. Through literature research and on-site exploration, specific characteristics of low street space quality in the offsite area can be determined: the houses on two sides are short and dense, the building layout is disordered, the outer vertical face of the building is old, and the pipeline layout is disordered; the road is narrow, people and vehicles are mixed, the road environment is sanitary, dirty and messy, the greening level is low, and the sky visibility is poor; a large number of shop boards or billboards with different styles and colors exist on part of streets, and small vendors gather the boards.
The street view images of four viewing angles are labeled according to the above characteristics, and are classified into two types, namely low-quality images and high-quality images (as shown in fig. 2). Images that satisfy one or more features are labeled as streets with low spatial quality, otherwise, as streets with high spatial quality. 200 street view images are marked at each of the four visual angles, wherein 100 street view image samples with low street space quality and 100 street view image samples with high street space quality are respectively marked at each visual angle, 70% of the street view images are selected as a training set of a street space quality evaluation model, and the rest 30% of the street view images are selected as a test set.
The method comprises the steps of obtaining a plurality of characteristics of a street view image sample, wherein the characteristics comprise manually designed characteristics (accelerated robust characteristics and directional gradient histogram characteristics) and characteristics (semantic characteristics) based on deep learning, combining a plurality of characteristic vectors in a serial mode, respectively constructing a plurality of street space quality evaluation models under each visual angle by using a Support Vector Machine (SVM) and a Random Forest algorithm (RF), measuring model performance through classification accuracy and Kappa coefficients of each model on a test set, and selecting an optimal model and a corresponding characteristic combination mode under each visual angle.
The SVM and the RF are used for judging the space quality of the street view image. The SVM converts the linear indivisible samples in the low-dimensional space into linear separable samples in the high-dimensional space, and searches for an optimal classification hyperplane in the high-dimensional space by the principle of interval maximization. In this embodiment, the SVM kernel function adopts a radial basis kernel function, and the regularization parameter C as a main parameter of the algorithm and the bandwidth gamma of the radial basis kernel function are screened by a grid search method.
And the RF utilizes a strategy of sample randomness and characteristic randomness to construct a plurality of decision trees which are independent of each other, and the generalization performance of the model is improved through a strategy of reducing the variance. The main parameters of the algorithm comprise the number of decision trees, the maximum depth of the decision trees and the available feature number of a single decision tree. In this embodiment, the number of decision trees for RF is 100, and the remaining parameters are determined by the grid search method.
Step three: and extracting the multi-class high-resolution remote sensing image features of the plots to be classified from the preprocessed high-resolution remote sensing images according to the characteristics of the plots in the rural areas in the city.
The characteristics of the villages in different regions on the high-resolution remote sensing image are different, and an index with good discrimination in one region is not applicable in another region. In this embodiment, the difference between the Guangzhou city village and the conventional city built-up area on the high-resolution remote sensing image is analyzed by combining the previous research and field investigation (as shown in table 1), a plurality of categories of features are obtained from the GF-2 image to depict the Guangzhou city village, including spectral features, shape features, texture features, building structures and scene features, and corresponding indexes are selected (as shown in table 2).
The spectral feature system comprises four original multispectral wave bands of a GF-2 image, a Mean value (Mean) and a Standard deviation (Standard deviation) of a first principal component PCA1 which contains most information of the image after dimensionality reduction, and 2 common indexes of Brightness (Brightness) and a maximum value (max-diff) of Brightness difference.
The Shape features include Area of land (Area), Density (Density), Shape index (Shape index)3 indices. The gray level co-occurrence matrix obtains the spatial correlation rule of the image gray level by researching the joint distribution condition of the gray level pixels, and is the most common texture extraction method in the remote sensing image. Texture features include correlation, entropy and standard deviation of Gray Level Co-occurrence Matrix (GLCM).
TABLE 1 difference between Guangzhou city village and conventional city built-up area in high resolution remote sensing image
TABLE 2 village-in-city feature system based on high-resolution remote sensing image and street view image
The architectural structural features include the area mean of the architectural objects and the standard deviation of the mean of the PCA1 of the architectural objects. In the scene characteristics, NDVI is adopted to represent vegetation coverage inside the plots, and Veg _ P is used to represent greening conditions of surrounding environments (vegetation area ratio in a buffer area of 15m in each plot). Shadow _ P represents the proportion of the Shadow area in each block to the area of the block, and the Shadow data is subjected to threshold segmentation extraction based on PCA1 of the high-resolution remote sensing image. As shown in table 2, 23 features were finally obtained from the GF-2 images.
Step four: and drawing the street space quality of the selected area by using the optimal street space quality evaluation model under each visual angle and a corresponding characteristic combination mode, and acquiring a spatially continuous street space quality distribution map of the selected area by using a spatial interpolation strategy so as to extract a plurality of street space quality indexes in the plots to be classified.
In this embodiment, based on the optimal street view image feature combination at each viewing angle, the street space quality in the street view image of the offsite 59720 is evaluated by using the optimal street space quality evaluation model at each viewing angle. The higher the probability value that a certain street view image belongs to a high-quality street is judged by the model, the higher the street space quality is, and the spatial distribution of the street space quality of the sampling point scale of the overseas area is finally obtained by calculating the mean value of the street space quality reflected by 4 street view images of each sampling point.
And (3) acquiring a spatially continuous street space quality distribution map by adopting an inverse distance weighted spatial interpolation strategy based on the spatial distribution of the street space quality of the sampling point scale, thereby acquiring 5 features of the mean value, the variance, the maximum value, the minimum value and the range of the street space quality in each block.
Step five: and performing heterogeneous fusion on the multi-class high-resolution remote sensing image features and the street quality features of each plot to be classified, constructing a classifier by using a machine learning algorithm, and judging whether the rest plots are villages in the city or not.
In this example, 8 rural and 23 non-rural plots are labeled as training sets in the area to be classified against GF-2 images, depending on the field investigation. The high-resolution remote sensing image and the street view image are heterogeneous data and are suitable for feature level fusion. And respectively based on the high-resolution remote sensing image features, the street view image features and the fused features of the high-resolution remote sensing image features and the street view image features, adopting a random forest construction classifier, and judging whether the rest plots are rural areas.
In summary, the flow of identifying villages in town provided by this embodiment is shown in fig. 3.
After the fifth step, the method for identifying villages in the city provided by this embodiment may further include:
step six: and comprehensively acquiring the urban-rural area spatial distribution condition of the plot scale by combining field investigation, Google Earth map and visual interpretation technology, taking the urban-rural area spatial distribution condition as a ground true value, and evaluating the model performance by utilizing the integral classification precision, the Kappa coefficient, the user precision and the producer precision of the urban-rural area plot.
In this embodiment, the current situation of spatial distribution of urban villages in overseas areas in the size of a parcel is comprehensively obtained by means of official publishing data, field research, Google Earth visual interpretation and the like, and the current situation is used as a Ground truth value (Ground truth) to verify the classification result. Table 3 shows the accuracy of extracting villages in cities based on various image features, and it can be seen that the model established based on various features of the high-resolution remote sensing image can obtain better classification accuracy, the overall accuracy can reach 94.5%, and the Kappa coefficient is 0.58; the precision (63.5%) of a model producer established based on the street view image is slightly better than that (63.1%) of a model producer based on the high-resolution remote sensing image, but the user precision, the overall precision and the Kappa coefficient are far lower than those of the other party.
TABLE 3 accuracy of urban village extraction based on various image characteristics
Note: the user precision and the producer precision are extracted results aiming at the urban rural and rural building areas;
user precision is 1-error division error; 1-missing error in producer precision
Although the overall accuracy of the model established based on the street space quality features is low, after the street space quality information is blended in the urban village extraction process based on the high-resolution remote sensing image, all evaluation indexes of the model are improved to different degrees. The user precision is improved to the highest extent, and compared with a model based on a high-resolution remote sensing image or a street view image, the user precision is improved by 15.5% and 47.4%, and the precision of a producer is only slightly improved. The model with the optimal performance of each measurement index is a model formed by fusing the characteristics of the high-resolution remote sensing image and the street view image.
Fig. 4 is a comparison graph of the comprehensive urban village extraction results based on different images, which further illustrates the reason for the accuracy improvement. The phenomenon of wrong division exists when the urban village is extracted based on the high-resolution remote sensing image, namely, part of low building dense areas are judged as the urban village, and the street space quality of the areas is medium and can be observed through the street view image, and the urban village is mainly a factory storehouse and other building areas and does not belong to the urban village area. And when the urban village is extracted based on the street view image, part of old districts or small vendor gathering areas can be identified as the urban village, and the sparse buildings or the high building floors in the areas can be observed through the high-resolution remote sensing image, so that the urban village does not belong to the urban village. The improvement of the urban village identification precision is mainly the improvement of the user precision, namely, the misclassification phenomenon of the urban village is reduced after the multi-source images are fused. The information provided by the high-resolution remote sensing image of the aerial view and the street view image of the human-based view can be comprehensively complemented, and the accuracy of identifying urban villages and rural villages is improved.
In this embodiment, the importance of each feature in the fused feature (high-resolution remote sensing image feature + street view image feature) model may also be obtained based on an RF algorithm.
In the embodiment, an average Gini reduction value (Mean reduction Gini) method is selected to evaluate the importance of the features, and the method is based on the principle that the Gini coefficient reduction value of the index is calculated when node segmentation is carried out according to the principle of minimum node purity, all decision trees are averaged after all nodes in RF are summed, and the average is used as the importance of the index, and the importance of each feature is analyzed based on the method.
As shown in table 4, the most important feature is the street space quality feature, and the minimum and mean indicators contribute about 20% of the feature importance. The third to fifth important features are texture feature (GLCM _ correction), scene feature (Veg _ p), and shape feature (Area) in this order. The 5 indices for measuring street space quality extracted based on street view images contribute 31.6% feature importance. Therefore, the street space quality plays a key role in extracting the urban villages.
TABLE 4 feature importance and ranking
Note: the importance of only some of the features is enumerated here.
In the embodiment, the overseas district in Guangzhou city is taken as an example, and the high-resolution remote sensing image and the street view image are fused to perform comprehensive extraction of villages and villages in the city. The extraction precision of the village in the city after the two types of images are fused can reach 96.1 percent, which is higher than that based on a single image, and the fact that the information of the high-resolution remote sensing image and the street view image are fused is proved to improve the identification precision of the village in the city.
The urban village is a complex urban landscape, roof building materials of the urban village are various, space composition of internal ground features is not obviously regular, and the urban village shows differences from conventional urban built areas in the aspects of physical landscapes such as spectral textures, building structures and the like in the high-resolution remote sensing image, so that a better urban village classification result can be obtained based on the characteristics of the high-resolution remote sensing image such as textures, scenes, shapes and the like. In addition, the urban village and the conventional urban built-up area have great difference on the social and economic level. In the full-feature model, the street space quality features acquired based on the street view images have the highest importance among 28 features, which shows that the difference of the social and economic information is also an important feature for distinguishing the rural area from the conventional built-up urban area. The high-resolution remote sensing image and the street view image are respectively long in describing the two types of features, and the high-resolution remote sensing image and the street view image can effectively integrate the features of different types after being fused to construct a feature space with higher discrimination, so that the inter-view image fusion improves the accuracy of identifying urban villages and rural villages.
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 and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A city village identification method fusing a high-resolution remote sensing image and a street view image is characterized by comprising the following steps:
the method comprises the following steps: acquiring a high-resolution remote sensing image of a selected area for preprocessing, extracting a region to be classified from the preprocessed high-resolution remote sensing image according to the land utilization condition of the selected area, and performing multi-scale segmentation on the region to be classified to obtain a plurality of plots to be classified;
step two: acquiring and labeling multi-view street view images of the selected area, extracting a plurality of street view image features, constructing a street space quality evaluation model on a training set by utilizing a machine learning algorithm, and selecting an optimal street space quality evaluation model and a corresponding optimal street view image feature combination mode under each view;
step three: extracting the characteristics of various high-resolution remote sensing images of the plots to be classified from the preprocessed high-resolution remote sensing images according to the characteristics of the plots in the rural areas in the city;
step four: drawing the street space quality of the selected area by using the optimal street space quality evaluation model under each visual angle and a corresponding characteristic combination mode, and acquiring a spatially continuous street space quality distribution map of the selected area by using a spatial interpolation strategy so as to extract a plurality of street space quality indexes in the plots to be classified;
step five: and performing heterogeneous fusion on the multi-class high-resolution remote sensing image features and the street quality features of each plot to be classified, constructing a classifier by using a machine learning algorithm, and judging whether the rest plots are villages in the city or not.
2. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 1, wherein the method comprises the following steps:
in the first step, the high-resolution remote sensing image comes from a GF-2PMS1 sensor and comprises a panchromatic wave band and a multispectral wave band, and the preprocessing step comprises the following steps: orthorectification, radiometric calibration, atmospheric rectification, image registration and image fusion.
3. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 2, wherein the method comprises the following steps:
in the first step, the land utilization conditions of the selected area comprise water, vegetation, roads, urban village building areas and non-urban village building areas, wherein the areas to be classified are the urban village areas and the non-urban village building areas.
4. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 3, wherein in the step one, the multi-scale segmentation of the region to be classified comprises the following steps:
and constructing a partition layer with two scales of a building object level and a land block level by taking the road data and the vegetation data as constraint conditions, and performing multi-scale partition on the to-be-classified area.
5. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 4, wherein the method comprises the following steps:
in the second step, the streetscape image features comprise an acceleration robust feature, a directional gradient histogram feature and a semantic feature, and the machine learning algorithm comprises a support vector machine and a random forest algorithm.
6. The method for identifying villages in cities and villages integrating high-resolution remote sensing images and street view images as claimed in claim 5, wherein in said step two, a street space quality evaluation model is computationally constructed on a training set by machine learning, and selecting an optimal street space quality evaluation model and a corresponding optimal street view image feature combination mode under each view angle comprises:
dividing the street view image features into a training set and a test set, constructing a plurality of street space quality evaluation models on the training set by using the machine learning algorithm based on the training set, measuring model performance by using the classification precision of each model on the test set and the Kappa coefficient, and selecting the optimal street space quality model and the corresponding optimal street view image feature combination mode under each view angle.
7. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 6, wherein:
in the third step, the categories of the multi-category high-resolution remote sensing image features comprise spectral features, shape features, texture features, building structures and scene features, and each category corresponds to a plurality of high-resolution remote sensing image features.
8. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 7, wherein:
in the fourth step, the street quality features are the mean, variance, maximum, minimum and range of street space quality within each parcel to be classified.
9. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 8, wherein:
in the fifth step, the machine learning algorithm is a random forest algorithm.
10. The method for identifying villages in cities by fusing high-resolution remote sensing images and street view images according to claim 9, wherein after the step five, the identification method further comprises the following steps:
step six: comprehensively acquiring the urban-rural spatial distribution condition of the plot scale as a ground real value by combining field research, a Google Earth map and a visual interpretation technology; and evaluating the performance of the model by utilizing the overall classification precision, the Kappa coefficient, the user precision of the urban village plot and the producer precision.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010481052.0A CN111598048B (en) | 2020-05-31 | 2020-05-31 | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010481052.0A CN111598048B (en) | 2020-05-31 | 2020-05-31 | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111598048A true CN111598048A (en) | 2020-08-28 |
| CN111598048B CN111598048B (en) | 2021-06-15 |
Family
ID=72192485
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010481052.0A Active CN111598048B (en) | 2020-05-31 | 2020-05-31 | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111598048B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329853A (en) * | 2020-11-05 | 2021-02-05 | 湖南省沅江市安沅水利水电建设有限公司 | Traffic accident detection method, device, equipment and medium based on image fusion |
| CN112347926A (en) * | 2020-11-06 | 2021-02-09 | 天津市勘察设计院集团有限公司 | High-resolution image urban village detection method based on building form distribution |
| CN112818747A (en) * | 2020-12-31 | 2021-05-18 | 上海应用技术大学 | Urban characteristic neighborhood population density estimation method and system method based on spatial big data |
| CN112967286A (en) * | 2021-05-19 | 2021-06-15 | 航天宏图信息技术股份有限公司 | Method and device for detecting newly added construction land |
| CN113469226A (en) * | 2021-06-16 | 2021-10-01 | 中国地质大学（武汉） | Street view image-based land utilization classification method and system |
| CN113610346A (en) * | 2021-07-02 | 2021-11-05 | 华南农业大学 | Village development potential evaluation and village classification method and device based on multi-source data |
| CN114202622A (en) * | 2022-02-18 | 2022-03-18 | 腾讯科技（深圳）有限公司 | Virtual building generation method, device, equipment and computer readable storage medium |
| CN117787813A (en) * | 2024-02-23 | 2024-03-29 | 中国农业科学院农业资源与农业区划研究所 | Fine-granularity feature constraint-based urban village identification method and device and electronic equipment |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279951A (en) * | 2013-05-13 | 2013-09-04 | 武汉理工大学 | Object-oriented remote sensing image building and shade extraction method of remote sensing image building |
| CN104105106A (en) * | 2014-07-23 | 2014-10-15 | 武汉飞脉科技有限责任公司 | Wireless communication network intelligent-antenna-covered scene automatic classification and recognition method |
| CN104463168A (en) * | 2014-11-25 | 2015-03-25 | 中国科学院地理科学与资源研究所 | Automatic waste empty house site information extraction method based on remote-sensing image |
| CN104537975A (en) * | 2015-01-16 | 2015-04-22 | 北京智谷睿拓技术服务有限公司 | Display control method and device and display equipment |
| US20180068206A1 (en) * | 2016-09-08 | 2018-03-08 | Mentor Graphics Corporation | Object recognition and classification using multiple sensor modalities |
| CN108629287A (en) * | 2018-04-09 | 2018-10-09 | 华南农业大学 | A kind of remote sensing image terrain classification method |
| CN108898096A (en) * | 2018-06-27 | 2018-11-27 | 重庆交通大学 | A kind of quick accurate extracting method of the information towards high score image |
| CN110263717A (en) * | 2019-06-21 | 2019-09-20 | 中国科学院地理科学与资源研究所 | It is a kind of incorporate streetscape image land used status determine method |
| CN110263716A (en) * | 2019-06-21 | 2019-09-20 | 中国科学院地理科学与资源研究所 | A kind of remote sensing image super-resolution windy and sandy soil drafting method based on street view image |
| CN110298321A (en) * | 2019-07-02 | 2019-10-01 | 中国科学院遥感与数字地球研究所 | Route denial information extraction based on deep learning image classification |
| US10503966B1 (en) * | 2018-10-11 | 2019-12-10 | Tindei Network Technology (Shanghai) Co., Ltd. | Binocular pedestrian detection system having dual-stream deep learning neural network and the methods of using the same |
| CN110726677A (en) * | 2019-10-18 | 2020-01-24 | 中国科学院地理科学与资源研究所 | Polluted site remote sensing detection and space hot area identification system and method |
| CN110852262A (en) * | 2019-11-11 | 2020-02-28 | 南京大学 | Agricultural land extraction method based on time sequence top-grade first remote sensing image |
| CN111047695A (en) * | 2019-12-03 | 2020-04-21 | 中国科学院地理科学与资源研究所 | Method for extracting height spatial information and contour line of urban group |
| CN111127493A (en) * | 2019-11-12 | 2020-05-08 | 中国矿业大学 | Remote sensing image semantic segmentation method based on attention multi-scale feature fusion |
-
2020
- 2020-05-31 CN CN202010481052.0A patent/CN111598048B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279951A (en) * | 2013-05-13 | 2013-09-04 | 武汉理工大学 | Object-oriented remote sensing image building and shade extraction method of remote sensing image building |
| CN104105106A (en) * | 2014-07-23 | 2014-10-15 | 武汉飞脉科技有限责任公司 | Wireless communication network intelligent-antenna-covered scene automatic classification and recognition method |
| CN104463168A (en) * | 2014-11-25 | 2015-03-25 | 中国科学院地理科学与资源研究所 | Automatic waste empty house site information extraction method based on remote-sensing image |
| CN104537975A (en) * | 2015-01-16 | 2015-04-22 | 北京智谷睿拓技术服务有限公司 | Display control method and device and display equipment |
| US20180068206A1 (en) * | 2016-09-08 | 2018-03-08 | Mentor Graphics Corporation | Object recognition and classification using multiple sensor modalities |
| CN108629287A (en) * | 2018-04-09 | 2018-10-09 | 华南农业大学 | A kind of remote sensing image terrain classification method |
| CN108898096A (en) * | 2018-06-27 | 2018-11-27 | 重庆交通大学 | A kind of quick accurate extracting method of the information towards high score image |
| US10503966B1 (en) * | 2018-10-11 | 2019-12-10 | Tindei Network Technology (Shanghai) Co., Ltd. | Binocular pedestrian detection system having dual-stream deep learning neural network and the methods of using the same |
| CN110263717A (en) * | 2019-06-21 | 2019-09-20 | 中国科学院地理科学与资源研究所 | It is a kind of incorporate streetscape image land used status determine method |
| CN110263716A (en) * | 2019-06-21 | 2019-09-20 | 中国科学院地理科学与资源研究所 | A kind of remote sensing image super-resolution windy and sandy soil drafting method based on street view image |
| CN110298321A (en) * | 2019-07-02 | 2019-10-01 | 中国科学院遥感与数字地球研究所 | Route denial information extraction based on deep learning image classification |
| CN110726677A (en) * | 2019-10-18 | 2020-01-24 | 中国科学院地理科学与资源研究所 | Polluted site remote sensing detection and space hot area identification system and method |
| CN110852262A (en) * | 2019-11-11 | 2020-02-28 | 南京大学 | Agricultural land extraction method based on time sequence top-grade first remote sensing image |
| CN111127493A (en) * | 2019-11-12 | 2020-05-08 | 中国矿业大学 | Remote sensing image semantic segmentation method based on attention multi-scale feature fusion |
| CN111047695A (en) * | 2019-12-03 | 2020-04-21 | 中国科学院地理科学与资源研究所 | Method for extracting height spatial information and contour line of urban group |
Non-Patent Citations (8)
| Title |
|---|
| CHENGMING ZHANG ET AL: "Improved remote sensing image classification based on multi-scale feature fusion", 《REMOTE SENSING》 * |
| L LIU ET AL: "A machine learning-based method for the large- scale evaluation of the qualities of the urban environmen", 《COMPUTERS ENVIRONMENT & URBAN SYSTEMS》 * |
| LIN, J ET AL: "Analyzing the spatial factors related to the distributions of building heights in urban areas: A comparative case study in Guangzhou and Shenzhen", 《SUSTAINABLE CITIES AND SOCIETY》 * |
| ZOU，TY ET AL: "Polarimetric SAR Image Classification Using Multifeatures Combination and Extremely Randomized Clustering Forests", 《EURASIP JOURNAL ON ADVANCES IN SIGNAL PROCESSING》 * |
| 叶宇等: "人本尺度的街道空间晶质测度 一一结合街景数据和亲归于析技术的大规模、高精度评价框架", 《国际城市规划》 * |
| 李科等: "基于多特征组合与优化BoW模型的影像分类技术研究", 《***仿真学报》 * |
| 赵云涵等: "耦合多源大数据提取城中村建筑物——以广州市天河区为例", 《地理与地理信息科学》 * |
| 赵崇悦: "基于多特征学习的遥感影像复原与分类", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329853A (en) * | 2020-11-05 | 2021-02-05 | 湖南省沅江市安沅水利水电建设有限公司 | Traffic accident detection method, device, equipment and medium based on image fusion |
| CN112347926A (en) * | 2020-11-06 | 2021-02-09 | 天津市勘察设计院集团有限公司 | High-resolution image urban village detection method based on building form distribution |
| CN112347926B (en) * | 2020-11-06 | 2023-05-23 | 天津市勘察设计院集团有限公司 | High-resolution image city village detection method based on building morphology distribution |
| CN112818747A (en) * | 2020-12-31 | 2021-05-18 | 上海应用技术大学 | Urban characteristic neighborhood population density estimation method and system method based on spatial big data |
| CN112967286A (en) * | 2021-05-19 | 2021-06-15 | 航天宏图信息技术股份有限公司 | Method and device for detecting newly added construction land |
| CN112967286B (en) * | 2021-05-19 | 2021-09-17 | 航天宏图信息技术股份有限公司 | Method and device for detecting newly added construction land |
| CN113469226A (en) * | 2021-06-16 | 2021-10-01 | 中国地质大学（武汉） | Street view image-based land utilization classification method and system |
| CN113610346A (en) * | 2021-07-02 | 2021-11-05 | 华南农业大学 | Village development potential evaluation and village classification method and device based on multi-source data |
| CN114202622A (en) * | 2022-02-18 | 2022-03-18 | 腾讯科技（深圳）有限公司 | Virtual building generation method, device, equipment and computer readable storage medium |
| CN117787813A (en) * | 2024-02-23 | 2024-03-29 | 中国农业科学院农业资源与农业区划研究所 | Fine-granularity feature constraint-based urban village identification method and device and electronic equipment |
| CN117787813B (en) * | 2024-02-23 | 2024-05-31 | 中国农业科学院农业资源与农业区划研究所 | Fine-granularity feature constraint-based urban village identification method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111598048B (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111598048B (en) | Urban village-in-village identification method integrating high-resolution remote sensing image and street view image | |
| Xu et al. | Classification of local climate zones using ASTER and Landsat data for high-density cities | |
| Herold et al. | Spatial metrics and image texture for mapping urban land use | |
| Heiden et al. | Urban structure type characterization using hyperspectral remote sensing and height information | |
| Huang et al. | Urban functional zone mapping by integrating high spatial resolution nighttime light and daytime multi-view imagery | |
| CN109919875B (en) | High-time-frequency remote sensing image feature-assisted residential area extraction and classification method | |
| Zhou et al. | Individual tree parameters estimation for plantation forests based on UAV oblique photography | |
| CN110309780A (en) | High resolution image houseclearing based on BFD-IGA-SVM model quickly supervises identification | |
| CN106501186B (en) | A kind of soil moisture content product NO emissions reduction method | |
| CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
| Chen et al. | Detecting changes in high-resolution satellite coastal imagery using an image object detection approach | |
| CN111242223B (en) | Street space quality evaluation method based on streetscape image multi-feature fusion | |
| CN109308451A (en) | A kind of high score data information extraction system and method | |
| Truax | Comparing spectral and object based approaches for classification and transportation feature extraction from high resolution multispectral imagery | |
| Lefebvre et al. | Monitoring the morphological transformation of Beijing old city using remote sensing texture analysis | |
| CN112800827A (en) | Hyperspectral image classification experimental method | |
| Zhang | A time-series approach to detect urbanized areas using biophysical indicators and landsat satellite imagery | |
| Xi et al. | High-Confidence Sample Generation Technology and Application for Global Land-Cover Classification | |
| Shao et al. | Mapping impervious surfaces with a hierarchical spectral mixture analysis incorporating endmember spatial distribution | |
| CN112347926B (en) | High-resolution image city village detection method based on building morphology distribution | |
| CN113468982A (en) | Method, device and storage medium for classifying urban functional areas | |
| Liu et al. | Automated extraction of urban roadside trees from mobile laser scanning point clouds based on a voxel growing method | |
| Forghani et al. | Extracting terrain categories from multi-source satellite imagery | |
| Bai et al. | Estimating the Aboveground Biomass of Urban Trees by Combining Optical and Lidar Data: A Case Study of Hengqin, Zhuhai, China | |
| LOTFIAN | Urban climate modeling: case study of Milan city |
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 |