CN114120133A - Urban road information extraction method for aggregating multiple factors - Google Patents
Urban road information extraction method for aggregating multiple factors Download PDFInfo
- Publication number
- CN114120133A CN114120133A CN202111466300.5A CN202111466300A CN114120133A CN 114120133 A CN114120133 A CN 114120133A CN 202111466300 A CN202111466300 A CN 202111466300A CN 114120133 A CN114120133 A CN 114120133A
- Authority
- CN
- China
- Prior art keywords
- optimal
- road information
- segmentation
- scale
- classification
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to an urban road information extraction method for aggregating multiple factors, which comprises the following steps: preprocessing the remote sensing image; constructing an optimal segmentation scale calculation model to calculate optimal scale based on homogeneity inside the objects and heterogeneity between adjacent objects; setting a shape factor and compactness; carrying out segmentation processing on the remote sensing image according to the optimal scale, the shape factor and the compactness; obtaining a classification rule based on a SEATH algorithm; and obtaining a road information extraction result according to the classification rule, and outputting the result. The invention provides a method for extracting urban road information of a high-resolution remote sensing image, which is different from the existing method for extracting urban road information of the high-resolution remote sensing image, is efficient and wide in application range, and the method adopts an optimal segmentation scale calculation model and a SEATH algorithm to quickly extract the urban road information.
Description
Technical Field
The invention relates to the technical field of remote sensing images, in particular to an urban road information extraction method capable of aggregating multiple factors.
Background
The high-resolution remote sensing image has the characteristics of high resolution, obvious geometric structure and textural features and the like, and is widely applied to extraction of various ground objects. With the continuous improvement of the resolution of the remote sensing image, the ground feature information in the image is more detailed, and further support can be provided for the identification and extraction of the artificial ground feature. As one of important artificial ground features, roads are not only main research objects of modern traffic systems, but also main objects recorded and extracted by map identification, Geographic Information Systems (GIS) and remote sensing technologies, and how to quickly and accurately extract urban road information from high-resolution remote sensing image data becomes one of the hotspots of related research.
At present, scholars at home and abroad make a great deal of research on the road extraction method of the high-resolution remote sensing image, and the method is divided into a medium-low level method and a medium-high level method according to the extraction technology. The method for extracting the road characteristics of the middle-low level comprises a road tracking method, a dynamic planning method, a Snake method, a multi-temporal analysis method, a stereopair analysis method and the like; and a middle-high level road feature extraction method knowledge expression method, a fuzzy modeling method and the like. From the research, it is easy to find that most of the current mature methods are directed at medium and low resolution remote sensing images, most of the high resolution remote sensing image road information extraction algorithms are modeling by using one or more characteristics of roads under a simple background, and topological characteristics and context semantic characteristics are not fully considered, so that the extraction accuracy of the roads is influenced. Some scholars who aim at the problems provide that the object-oriented technology is applied to information extraction of high-resolution remote sensing images, the noise problem existing in remote sensing classification can be effectively solved, context semantic information of object features is fully utilized, image information of ground objects is extracted in a corresponding scale by combining types of the ground objects, and therefore extraction accuracy of the image information is improved.
From a plurality of research results, the current research aiming at the road extraction of the high-resolution remote sensing image is divided into two types: the first type is that road information extraction precision is improved by optimizing segmentation scale; the current object-oriented optimal scale selection method mainly comprises three categories: (1) the method is simple and easy to implement, has certain subjectivity, and is difficult to determine whether the obtained result is optimal or not; (2) index evaluation method, the selection of the index is difficult to determine, and ideal results cannot be achieved; (3) the optimal segmentation scale is calculated by utilizing an optimal segmentation scale calculation model, and the segmentation quality of the image is judged according to the internal homogeneity and the heterogeneity of the objects, so that the method is objective and practical. In the second category, the extraction rule is optimized by selecting a reasonable algorithm, so that the extraction precision is improved; at present, a plurality of methods for establishing a road information extraction rule are available, such as a manual test method, a SEATH algorithm, a CART algorithm and the like. The SEATH algorithm can automatically select features and determine threshold values, and is embedded into the ecognition software, so that representative features and threshold values of the representative features can be conveniently and quickly obtained. However, both studies have a certain one-sidedness, and a complete and systematic road information extraction algorithm cannot be formed.
In view of the above disadvantages, it is desirable to provide a method for extracting urban road information by aggregating multiple factors, so as to implement complete and systematic extraction of urban road information.
Disclosure of Invention
The invention aims to solve the technical problems that in the prior art, urban road information extraction has one-sidedness, a complete and systematic road information extraction algorithm cannot be formed, and aiming at the defects in the prior art, the invention provides a multi-factor aggregated urban road information extraction method which adopts an optimal segmentation scale calculation model and a SEATH algorithm to rapidly extract urban road information.
In order to solve the technical problem, the invention provides a method for extracting urban road information by aggregating multiple factors, which comprises the following steps: preprocessing the remote sensing image; constructing an optimal segmentation scale calculation model to calculate optimal scale based on homogeneity inside the objects and heterogeneity between adjacent objects; setting a shape factor and compactness; carrying out segmentation processing on the remote sensing image according to the optimal scale, the shape factor and the compactness; obtaining a classification rule based on a SEATH algorithm; and obtaining a road information extraction result according to the classification rule, and outputting the result.
Preferably, in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on homogeneity inside the object and heterogeneity between adjacent objects, the homogeneity calculation is performed by setting a pre-segmented image to have m bands, and the weight given to each band when performing multi-scale segmentation is ti(i ═ 1, 2, …, m), homogeneity is expressed using a weighted sum of standard deviations across the various bands inside the subject, as follows:
in the formula, vb(b 1, 2.. said., m) denotes the homogeneity of the object over the band b, tb(b 1, 2.. m) represents the weight of the band b, viIs the standard deviation of the object i over the band b, aiDenotes the area of the object i, and n is the total number of objects obtained by dividing the entire region.
Preferably, in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on homogeneity inside the object and heterogeneity between adjacent objects, the heterogeneity is expressed by a weighted sum of standard deviations between the objects at each wavelength band,
in the formula Ib(b 1, 2.. said., m) denotes the object heterogeneity in the band b, tb(b 1, 2.. m) represents the weight of band b, wijRepresenting the adjacency relation between the object i and the object j, and if the object i and the object j are adjacent, wij1, otherwise wij＝0，yiIs the spectral average of object i, yjIs the average of the spectra of the object j,
the average of the spectra of the whole image.
Preferably, the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on homogeneity inside the object and heterogeneity between adjacent objects further comprises the following processes,
based on the homogeneity inside the object and the heterogeneity between adjacent objects, an improved remote sensing image segmentation quality evaluation function is provided,
F(v，I)＝(1-p)F(v)+pF(I)，
wherein v represents the weighted sum of the standard deviations on each wave band inside the object, I represents the weighted sum of the standard deviations on each wave band between the objects, p is the proportion of the I index in the objective function, F (v) is the homogeneity inside the object, F (I) is the heterogeneity between adjacent objects, and F (v, I) is the segmentation quality evaluation function;
constructing a segmentation quality function with a segmentation scale x as a variable through an interpolation function, and performing n +1 segmentation experiments on an image to be processed to obtain n + 1F (v, I) values;
determining the coefficient a0，a1，...，anTo obtain an optimal segmentation scale calculation model,
H(x)＝a0+a1x+a2x2+…+anxn；
and calculating the value of the segmentation scale x when the segmentation quality value is maximum through the optimal segmentation scale calculation model, wherein the value is the optimal scale.
Preferably, the shape factor is a parameter reflecting the shape integrity of the object, and the value is 0.5; the compactness describes the similarity degree of the object and the rectangle, and the value is 0.2.
Preferably, in the step of obtaining the classification rule based on the SEaTH algorithm, the optimal classification feature is screened based on the SEaTH algorithm, the feature threshold is determined, and the classification rule is obtained by establishing a rule set according to the optimal classification feature and the feature threshold.
Preferably, in the process of screening the optimal classification features and determining the feature threshold based on the SEATH algorithm, the method comprises the following steps,
the separation degree of the two categories on the separation characteristic is calculated by adopting the J-M distance, the distance J is calculated by the following method,
J＝2(1-e-B)，
wherein B represents the Papanicolaou distance, m1And m2Mean value, σ, representing the separation characteristics of two classes1And σ2A standard deviation of the separation characteristic representing the two classes;
selecting the first two separation features with the maximum separation degree as the optimal classification features, wherein the optimal classification features are used for classification;
the optimal threshold for the classification features of both classes is calculated.
Preferably, if the samples of both classes are subject to a normal distribution, the optimal threshold is calculated as follows,
in the formula, n1And n2Indicating the number of samples in both categories.
If the samples of the two classes do not follow a normal distribution, preferably, the optimal threshold is calculated as follows,
when J is more than 0.5 and less than 1.25, T ═ m2；
When J < 1.75, T ═ m (T + m)2)/2；
When J > 1.75, T' ═ T;
in the formula, n1And n2The number of samples representing the two categories,
preferably, the separation features include spectral features, shape features, texture features, custom features.
The implementation of the method for extracting the urban road information by aggregating the multiple factors has the following beneficial effects: (1) the urban road information extraction method based on the aggregation of the multiple factors is an efficient urban road information extraction method based on the high-resolution remote sensing image, is an efficient and automatic method, and has the advantages that the segmentation scale is obtained to be the optimal scale, the classification characteristic is optimal, the extraction precision is high, and the classification quality is good. (2) The method is wide in application range, is suitable for extracting the road information of the remote sensing images of the large, medium and small cities of different types, and has good extraction effect on the medium and small cities and good extraction effect on the large cities in the road extraction of the large, medium and small cities. (3) The method has the advantages that the extraction accuracy of roads in large, medium and small cities is continuously increased, but the classification quality is good, and the road extraction accuracy of the large, medium and small cities is that the large city is less than the medium city and less than the small city under the condition that the segmentation scale, the classification characteristics and the threshold value thereof are optimal.
Drawings
FIG. 1 is a flow chart of a method for extracting aggregated multi-factor urban road information according to an embodiment of the present invention;
FIG. 2 is a flowchart of the steps of selecting the optimal scale in the method for extracting aggregated multi-factor urban road information according to the embodiment of the present invention;
fig. 3 is a flowchart of steps of rule set establishment in the method for extracting aggregated multi-factor urban road information according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
FIG. 1 is a flow chart of a method for extracting aggregated multi-factor urban road information according to an embodiment of the present invention; as shown in fig. 1, the method for extracting aggregated multi-factor urban road information according to the embodiment of the present invention includes the following steps:
step S01: preprocessing the remote sensing image;
step S02: constructing an optimal segmentation scale calculation model to calculate optimal scale based on homogeneity inside the objects and heterogeneity between adjacent objects;
step S03: setting a shape factor and compactness;
step S04: carrying out segmentation processing on the remote sensing image according to the optimal scale, the shape factor and the compactness;
step S05: obtaining a classification rule based on a SEATH algorithm;
step S06: and obtaining a road information extraction result according to the classification rule, and outputting the result.
In the method for extracting urban road information by aggregating multiple factors, the data mainly used for research comprises remote sensing image data and road network data. Remote sensing image data adopts GF-2 image data as research data (4 m multispectral image and 1 m panchromatic image); the road network data is from a resource and environment scientific data center of China academy of sciences, has higher precision, and is widely applied to scientific research related to China.
The embodiment of the invention provides a multi-factor, systematic and high-resolution remote sensing image-based urban road information high-efficiency extraction method.
In the method for extracting the urban road information by aggregating the multiple factors, the remote sensing image is preprocessed by radiometric calibration, atmospheric correction, image fusion and the like, so that the road information can be highlighted, other information can be inhibited, and the accuracy of image segmentation and extraction is improved.
FIG. 2 is a flowchart of the steps of selecting the optimal scale in the method for extracting aggregated multi-factor urban road information according to the embodiment of the present invention; as shown in fig. 2, the optimal scale selection (i.e. optimal segmentation scale selection) includes the following steps:
step S11: performing image processing
Step S121 and step S122: determining intra-subject homogeneity and determining inter-subject heterogeneity;
step S13: obtaining an optimal segmentation scale calculation model;
step S14: and determining an optimal scale.
The multispectral remote sensing image is an image obtained by scanning a ground object reflected wave by a multiband scanner, and in order to improve the extraction accuracy of image information, the characteristics of each band of the remote sensing image should be comprehensively considered and fully utilized in the information extraction. The most ideal result of remote sensing image segmentation is as follows: good homogeneity inside the objects and good heterogeneity between adjacent objects can be embodied in each waveband. Therefore, the internal homogeneity and the inter-object heterogeneity of the objects can be used for evaluating the image quality segmentation effect, and an optimal segmentation scale calculation model is constructed to calculate the optimal scale.
In the method for extracting information of an urban road aggregating multi-factors provided by the embodiment of the invention, in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on the homogeneity inside the object and the heterogeneity between adjacent objects, a pre-segmented image is set to have m wave bands in the homogeneity calculation, and the weight given to each wave band when multi-scale segmentation is executed is ti(i＝1，2，…，m), homogeneity is expressed using a weighted sum of standard deviations across various bands inside the subject, as follows:
in the formula, vb(b 1, 2.. said., m) denotes the homogeneity of the object over the band b, tb(b 1, 2.. m) represents the weight of the band b, viIs the standard deviation of the object i over the band b, aiRepresenting the area of an object i, wherein n is the total number of the objects after the whole region is divided; the smaller the v value, the less the internal heterogeneity of the image object, and the better the homogeneity of the object.
In the method for extracting information of aggregated multi-factor urban roads provided by the embodiment of the invention, in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on the homogeneity inside the objects and the heterogeneity between adjacent objects, in the calculation of the heterogeneity, the weighted sum of the standard deviations of the objects on each wave band is adopted to express the heterogeneity,
in the formula Ib(b 1, 2.. said., m) denotes the object heterogeneity in the band b, tb(b 1, 2.. m) represents the weight of band b, wijRepresenting the adjacency relation between the object i and the object j, and if the object i and the object j are adjacent, wij1, otherwise wij＝0，yiIs the spectral average of object i, yjIs the average of the spectra of the object j,
the average of the spectra of the whole image.
In the method for extracting urban road information by aggregating multiple factors provided by the embodiment of the invention, the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on the homogeneity inside the object and the heterogeneity between adjacent objects further comprises the following steps:
based on the homogeneity inside the object and the heterogeneity between adjacent objects, an improved remote sensing image segmentation quality evaluation function is provided,
F(v，I)＝(1-p)F(v)+pF(I)，
wherein v represents the weighted sum of the standard deviations on each wave band inside the object, I represents the weighted sum of the standard deviations on each wave band between the objects, p is the proportion of the I index in the objective function, F (v) is the homogeneity inside the object, F (I) is the heterogeneity between adjacent objects, and F (v, I) is the segmentation quality evaluation function;
constructing a segmentation quality function with a segmentation scale x as a variable through an interpolation function, and performing n +1 segmentation experiments on an image to be processed to obtain n + 1F (v, I) values;
determining the coefficient a0，a1，...，anTo obtain an optimal segmentation scale calculation model,
H(x)＝a0+a1x+a2x2+…+anxn；
and calculating the value of the segmentation scale x when the segmentation quality value is maximum through the optimal segmentation scale calculation model, wherein the value is the optimal scale.
The analysis shows that: 1. v is increased along with the increase of the segmentation scale, the larger the segmentation scale is, the fewer the number of objects generated by segmenting the image is, the larger the area contained by a single object polygon is, the internal homogeneity is reduced, and the standard deviation is increased. 2. I decreases with increasing segmentation scale. The larger the segmentation scale is, the fewer the number of objects generated by segmenting the image is, the larger the area contained by a single object polygon is, the less the correlation between the objects is, and the more the heterogeneity is increased.
In the method for extracting aggregated multi-factor urban road information provided by the embodiment of the invention, the Shape factor (Shape) and the Compactness (Compactness) influence the quality of image segmentation, wherein the Shape factor is a parameter reflecting the Shape integrity of an object, and the Compactness can describe the similarity between the object and a rectangle. From a number of experiments it follows that both can be set to 0.5, 0.2 for road extraction, respectively.
Fig. 3 is a flowchart of the steps of rule set establishment in the method for extracting urban road information by aggregating multiple factors according to the embodiment of the present invention, and as shown in fig. 3, the rule set establishment includes the following steps:
step S21: creating a training set according to the optimal scale;
step S22: calculating the separation degree of the two categories on the separation characteristic by adopting the J-M distance;
step S23: selecting the first two separation features with the maximum separation degree as the optimal classification features;
step S24: determining an optimal threshold;
step S25: and outputting the rule set.
Image segmentation divides the image into meaningful separate regions, forming the primary image object, which becomes the basis for road interpretation. The key to the quality of road extraction is whether the rules in the decision tree can effectively distinguish various roads, so that the characteristics of various roads in the image must be fully known, and classification rules are established through the characteristic analysis of the roads.
In the method for extracting the aggregated multi-factor urban road information provided by the embodiment of the invention, the separation degree of two categories on the separation characteristic is calculated by adopting the J-M distance, the calculation method of the distance J is as follows,
J＝2(1-e-B)，
wherein B represents the Papanicolaou distance, m1And m2Mean value, σ, representing the separation characteristics of two classes1And σ2A standard deviation of the separation characteristic representing the two classes;
selecting the first two separation features with the maximum separation degree as the optimal classification features, wherein the optimal classification features are used for classification;
the optimal threshold for the classification features of both classes is calculated.
In the method for extracting urban road information by aggregating multiple factors provided by the embodiment of the invention, in the process of calculating the optimal threshold values of the classification features of the two categories, if the samples of the two categories are subjected to normal distribution, the optimal threshold values are calculated as follows,
in the formula, n1And n2Indicating the number of samples in both categories.
If the samples of the two classes do not follow a normal distribution, the optimal threshold is calculated as follows:
when J is more than 0.5 and less than 1.25, T' is m 2;
when 1.25 < J < 1.75, T ═ T + m 2)/2;
when J > 1.75, T' ═ T.
The SEaTH algorithm uses the degree of separation to evaluate the degree of discrimination between two classes on a feature. The separation degree is calculated by using a J-M (Jeffries-Matusita) distance, the distance is in a value range of [0, 2], 0 represents that two categories are almost completely mixed on a certain characteristic, and 2 represents that the two categories can be completely separated on a certain characteristic, however, the situation that J is 2 is rare in practical application, and some overlapping exists between the categories. In general, especially when a classification model is to be used for other image data, it is sufficient to select the first several features with the greatest separation for classification. Considering the portability of the classification model, the number of typical features used in classification is preferably minimized, so that only the top 2-bit features of the J value are usually reserved for classification.
Due to the high specificity of the high-resolution image, many features can be applied to establish a rule set for road extraction, however, in the case of limited samples, too many features can reduce the classification accuracy. Therefore, the SEaTH algorithm is required to choose the determination of objective and representative classification features and their thresholds.
The selection of the road classification characteristics should be representative and universal, and the road can be well extracted. In the method for extracting the aggregated multi-factor urban road information, provided by the embodiment of the invention, the separation characteristics comprise spectral characteristics, shape characteristics, texture characteristics and custom characteristics.
Wherein the spectral features include: (1) the object spectrum band mean value is characterized in that the reflectivity of the object region is reflected by the band mean value, and each band of the spectrum can be set with a threshold value to distinguish different road types and roads from other ground objects; (2) the standard deviation of a certain waveband of the object region is equal to the square root of the quotient obtained by subtracting the square sum of the mean value of the waveband from all pixel values in the object region and then dividing the sum by 1.
The shape features include: (1) the length-width ratio and the road have obvious linear characteristics, other objects can be removed through the linear characteristics, and the road is separated from the background. The aspect ratio index is a commonly used linear feature defined as the ratio of the length to the width of the smallest bounding rectangle of the object; (2) the compactness is the compactness of the object, and is calculated by the ratio of the minimum circumscribed rectangle of the object to the perimeter of the object, wherein the larger the ratio is, the higher the compactness of the object is.
The texture features include: (1) homogeneity (GLCM _ HOMOG) of the gray level co-occurrence matrix, wherein the homogeneity can describe the size of local texture change of the image, and if the image is locally and uniformly changed, the value of the homogeneity is larger; (2) the gray level co-occurrence matrix angular second moment (GLDV _ ANG _2) mainly analyzes the uniformity of image gray level distribution, calculates the sum of squares of elements in two directions, wherein the elements are perpendicular to each other, on a main diagonal line along the direction of the main diagonal line, respectively counts the sum of squares of the elements on the main diagonal line in two directions, when most of the values in the GLCM are distributed near the main diagonal line, the value of the gray level co-occurrence matrix angular second moment is larger, and if the distribution is uniform, the value is smaller.
The custom features mainly include a vegetation normalization index NDVI.
The method provided by the embodiment of the invention is adopted to carry out experiments and analysis, and particularly, the road information extraction is carried out on small cities (Ruichang), medium cities (Changsha) and large cities (Beijing). The high-resolution remote sensing image road information extraction experiment comprises the following steps: firstly, selecting the optimal scale of image data, wherein the shape factor is 0.5, the compactness is 0.2, and the spectral weight of each wave band is 1: 1 to obtain the optimal scale; secondly, performing multi-scale segmentation on image data in ecognition software according to the optimal segmentation scale to obtain an image object, selecting a representative object as a sample, outputting sample characteristic information, and determining a road classification rule and an optimal classification characteristic threshold; thirdly, classifying according to the optimal classification characteristics and the threshold value thereof to obtain a road information extraction result and output the result; and finally, carrying out precision evaluation on the classification result based on the vector sample points.
The results of the experiments and analyses were as follows:
the road extraction of an experiment I and a small city (Ruichang) is characterized in that an experimental research area is located in the Ruichang city district of partial west in the north of Jiangxi province, the road types of the area are mainly divided into an asphalt concrete road and a cement concrete road, the distribution is uniform, and the area has certain representativeness in the extraction of high-resolution image road information. Performing object-oriented road information extraction on the image according to the result of the optimal scale selection and the road classification rule selection, wherein the extracted road basically covers the main roads of the city in Ruichang, the cement concrete road and the asphalt concrete road are similar in quantity but different in distribution, the asphalt road is mainly distributed in the center of the city, and the houses are dense; cement concrete roads are mainly distributed around cities and in areas with relatively few houses. Because of the phenomena of partial wrong division and missed division caused by the interference of houses and the like, the experimental precision evaluation needs to be carried out on the extraction experiment. The production precision of the asphalt road is 0.78, namely the consistency degree between the actual classification and the reference classification reaches 78%, the missing classification error is 0.22, the user precision is 0.83, and the wrong classification error is 0.17; the production precision of the cement concrete road is 0.63, namely the consistency degree between the actual classification and the reference classification reaches 63%, the missing classification error is 0.37, the user precision is 0.99, and the wrong classification error is 0.01; the other production precision is 0.97, namely the consistency degree between the actual classification and the reference class reaches 97%, the missing classification error is 0.03, the user precision is 0.78, and the wrong classification error is 0.22; therefore, the asphalt road missing and wrong division errors are large, the cement concrete road missing and wrong division errors are large, the reason is related to the distribution of the asphalt road missing and wrong division errors, the influence of factors such as houses and the like and the defects of the classification method, the overall precision of the experiment is 0.83, the Kappa coefficient is 0.72, and the classification quality is good.
Experiment two, the road extraction of medium-sized city (Changsha), the urban area of Changsha city is basically asphalt concrete pavement, the cement concrete pavement is only distributed in the north bank where the double river ways and the fishing knife river intersect and the southeast of Changsha city, the road grade is not high, the width of the pavement is small, and the identification in the image is difficult. By adopting the method provided by the embodiment of the invention, the urban expressway in the Changsha city and the expressway in and out of the Changsha city have better extraction effect, the urban main road has better extraction effect, but the road can be disconnected, the cement concrete pavement has general extraction effect, the road information extraction effect is better on the whole, and the wrong separation and the missing separation still exist. The production precision of the asphalt road is 0.66, namely the consistency degree between the actual classification and the reference classification reaches 66%, the missing classification error is 0.34, the user precision is 0.98, and the wrong classification error is 0.02; the production precision of the cement concrete road is 0.19, namely the consistency degree between the actual classification and the reference classification reaches 19 percent, the missing classification error is 0.81, the user precision is 1, and the wrong classification error is 0; the other production precision is 0.98, namely the consistency degree between the actual classification and the reference class reaches 98 percent, the missing classification error is 0.02, the user precision is 0.69, and the wrong classification error is 0.31; therefore, the asphalt road is large in missing marks, the cement concrete road is large in missing mark errors, other wrong mark errors are large, the overall precision of the experiment is 0.80, the Kappa coefficient is 0.61, and the classification quality is good.
Experiment III, extraction of roads in a large city (Beijing), wherein the roads within five rings of the Beijing city basically realize full coverage of asphalt roads, so that cement concrete pavements do not exist within five rings, and the classification of the roads is only asphalt roads and other roads. By adopting the method provided by the embodiment of the invention, the urban expressway (loop) extracted from the road in Beijing City and the expressway in and out of Beijing City have better extraction effect and more complete road extraction, and the extraction of the roads such as the urban arterial road and the like is intermittent, and some roads are missed and some houses are wrongly divided into roads. The production precision of the asphalt road is 0.66, namely the consistency degree between the actual classification and the reference classification reaches 66%, the missing classification error is 0.34, the user precision is 0.98, and the wrong classification error is 0.02; the other production precision is 0.98, namely the consistency degree between the actual classification and the reference class reaches 98 percent, the missing classification error is 0.02, the user precision is 0.65, and the wrong classification error is 0.35; therefore, the missing separation errors of the asphalt road are large, other wrong separation errors are large, the overall precision of the experiment is 0.78, the Kappa coefficient is 0.59, and the classification quality is good.
According to classification results and classification quality of Beijing, Changsha and Ruichang cities, Kappa coefficients of three types of cities of big, medium and small are respectively 0.59, 0.61 and 0.72, so that the method for extracting the urban road information by aggregating the multiple factors has wide application range, and the size of the city is in inverse proportion to extraction precision. The overall accuracy of the large, medium and small cities is 0.78, 0.80 and 0.83 respectively, the Kappa coefficients are 0.59, 0.61 and 0.72 respectively, the missing and wrong distribution errors of the asphalt road in the small city are large, the missing and wrong distribution errors of the cement concrete road are large, other wrong distribution errors are large, the missing distribution errors of the asphalt road in the medium city are large, the missing distribution errors of the cement concrete road are large, other wrong distribution errors are large, the missing distribution errors of the asphalt road in the large city are large, and other wrong distribution errors are large. Therefore, the method has higher classification precision in three types of cities, and the large-scale city is smaller than the medium-scale city and smaller than the small-scale city, but the Kappa coefficients of the method do not exceed 0.80, namely the classification quality does not reach an excellent degree, and the extraction precision of the method can be improved by reducing the missing classification errors and other wrong classification errors of the asphalt road and the cement concrete road.
In summary, the method for extracting the urban road information by aggregating the multiple factors has the following beneficial effects: the road information extraction method comprehensively considers the spectral information and the spatial information, adds the knowledge of people on object cognition, is matched with the process of people in object cognition to a certain extent, improves the information extraction accuracy rate, and effectively avoids the influence caused by noise. The core of the road information extraction method provided by the invention is selection of an optimal segmentation scale and establishment of a rule set. The core of the optimal segmentation scale selection is an optimal scale calculation model, and the core of the rule set is an SEATH algorithm.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for extracting urban road information aggregating multiple factors is characterized by comprising the following steps:
preprocessing the remote sensing image;
constructing an optimal segmentation scale calculation model to calculate optimal scale based on homogeneity inside the objects and heterogeneity between adjacent objects;
setting a shape factor and compactness;
carrying out segmentation processing on the remote sensing image according to the optimal scale, the shape factor and the compactness;
obtaining a classification rule based on a SEATH algorithm;
and obtaining a road information extraction result according to the classification rule, and outputting the result.
2. The method of extracting aggregated multifactor urban road information according to claim 1, characterized in that: in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on homogeneity inside the object and heterogeneity between adjacent objects, the homogeneity calculation is performed by setting a pre-segmented image to have m bands, and the weight given to each band when performing multi-scale segmentation is ti(i ═ 1, 2, …, m), the homogeneity is expressed as a weighted sum of the standard deviations across the various bands inside the subject, as follows:
in the formula, vb(b 1, 2.. said., m) denotes the homogeneity of the object over the band b, tb(b 1, 2.. m) represents the weight of the band b, viIs the standard deviation of the object i over the band b, aiDenotes the area of the object i, and n is the total number of objects obtained by dividing the entire region.
3. The method of extracting aggregated multifactor urban road information according to claim 1, characterized in that: in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on the homogeneity inside the objects and the heterogeneity between adjacent objects, the heterogeneity is expressed by adopting the weighted sum of the standard deviations of the objects on each wave band in the calculation of the heterogeneity,
in the formula Ib(b 1, 2.. said., m) denotes the object heterogeneity in the band b, tb(b 1, 2.. m) represents the weight of band b, wijRepresenting the adjacency relation between the object i and the object j, and if the object i and the object j are adjacent, wij1, otherwise wij＝0，yiIs the spectral average of object i, yiAs the target, the spectral average of j,
the average of the spectra of the whole image.
4. The method for extracting information on urban roads based on aggregation of multifactor according to claim 1, wherein in the step of constructing an optimal segmentation scale calculation model to calculate the optimal scale based on homogeneity inside the object and heterogeneity between adjacent objects, the method further comprises the following steps:
based on the homogeneity inside the object and the heterogeneity between adjacent objects, an improved remote sensing image segmentation quality evaluation function is provided,
F(v，I)＝(1-p)F(v)+pF(I)，
wherein v represents the weighted sum of the standard deviations on each wave band inside the object, I represents the weighted sum of the standard deviations on each wave band between the objects, p is the proportion of the I index in the objective function, F (v) is the homogeneity inside the object, F (I) is the heterogeneity between adjacent objects, and F (v, I) is the segmentation quality evaluation function;
constructing a segmentation quality function with a segmentation scale x as a variable through an interpolation function, and performing n +1 segmentation experiments on an image to be processed to obtain n + 1F (v, I) values;
determining the coefficient a0，a1，...，anTo obtain an optimal segmentation scale calculation model,
H(x)＝a0+a1x+a2x2+…+anxn；
and calculating the value of the segmentation scale x when the segmentation quality value is maximum through the optimal segmentation scale calculation model, wherein the value is the optimal scale.
5. The method of extracting aggregated multifactor urban road information according to claim 1, characterized in that: the shape factor is a parameter reflecting the shape integrity of the object, and the value is 0.5; and the compactness describes the similarity of the object and the rectangle, and the value is 0.2.
6. The method for extracting aggregated multifactor urban road information according to any one of claims 1 to 5, characterized in that: in the step of obtaining the classification rule based on the SEATH algorithm, the optimal classification feature is screened based on the SEATH algorithm, a feature threshold is determined, and a rule set is established according to the optimal classification feature and the feature threshold to obtain the classification rule.
7. The method for extracting information of aggregated multi-factor urban roads according to claim 6, wherein in the process of screening optimal classification features and determining feature threshold based on SEATH algorithm, the method comprises the following steps:
the separation degree of the two categories on the separation characteristic is calculated by adopting the J-M distance, the distance J is calculated by the following method,
J＝2(1-e-B)，
wherein B represents the Papanicolaou distance, m1And m2Mean value, σ, representing the separation characteristics of two classes1And σ2A standard deviation of the separation characteristic representing the two classes;
selecting the first two separation features with the maximum separation degree as the optimal classification features, wherein the optimal classification features are used for classification;
the optimal threshold for the classification features of both classes is calculated.
8. The method of extracting aggregated multifactor urban road information according to claim 7, characterized in that: if both classes of samples follow a normal distribution, the optimal threshold is calculated as follows,
in the formula, n1And n2Indicating the number of samples in both categories.
9. The method of extracting aggregated multifactor urban road information according to claim 7, characterized in that: if the samples of the two classes do not follow a normal distribution, the optimal threshold is calculated as follows,
when J is more than 0.5 and less than 1.25, T ═ m2；
When J < 1.75, T ═ m (T + m)2)/2；
When J > 1.75, T' ═ T;
in the formula, n1And n2The number of samples representing the two categories,
10. the method of extracting aggregated multifactor urban road information according to claim 7, characterized in that: the separation characteristics comprise spectral characteristics, shape characteristics, texture characteristics and custom characteristics.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111466300.5A CN114120133A (en) | 2021-12-03 | 2021-12-03 | Urban road information extraction method for aggregating multiple factors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111466300.5A CN114120133A (en) | 2021-12-03 | 2021-12-03 | Urban road information extraction method for aggregating multiple factors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114120133A true CN114120133A (en) | 2022-03-01 |
Family
ID=80366022
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111466300.5A Pending CN114120133A (en) | 2021-12-03 | 2021-12-03 | Urban road information extraction method for aggregating multiple factors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114120133A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100272357A1 (en) * | 2006-07-28 | 2010-10-28 | University Of New Brunswick | Method of image segmentation |
| CN104899897A (en) * | 2015-05-27 | 2015-09-09 | 中国科学院地理科学与资源研究所 | High-resolution remote-sensing image land cover change detection method based on history data mining |
| CN107609526A (en) * | 2017-09-21 | 2018-01-19 | 吉林大学 | Rule-based fine dimension city impervious surface rapid extracting method |
| CN109580512A (en) * | 2018-12-10 | 2019-04-05 | 北京航天泰坦科技股份有限公司 | A kind of wheat nitrogen nutritional surveillance method based on spectrum and textural characteristics |
-
2021
- 2021-12-03 CN CN202111466300.5A patent/CN114120133A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100272357A1 (en) * | 2006-07-28 | 2010-10-28 | University Of New Brunswick | Method of image segmentation |
| CN104899897A (en) * | 2015-05-27 | 2015-09-09 | 中国科学院地理科学与资源研究所 | High-resolution remote-sensing image land cover change detection method based on history data mining |
| CN107609526A (en) * | 2017-09-21 | 2018-01-19 | 吉林大学 | Rule-based fine dimension city impervious surface rapid extracting method |
| CN109580512A (en) * | 2018-12-10 | 2019-04-05 | 北京航天泰坦科技股份有限公司 | A kind of wheat nitrogen nutritional surveillance method based on spectrum and textural characteristics |
Non-Patent Citations (6)
| Title |
|---|
| PRASHANTH R. MARPU 等: "A procedure for automatic object-based classification", 《RESEARCHGATE》 * |
| 何敏 等: "面向对象的最优分割尺度计算模型", 《大地测量与地球动力学》 * |
| 李宏力等: "基于资源三号卫星影像的面向对象分类方法的关键技术研究", 《测绘与空间地理信息》 * |
| 章程等: "基于最优尺度模型的建筑物信息提取", 《贵州大学学报(自然科学版)》 * |
| 郑东玉 等: "面向对象建筑物目标提取的最优分割尺度选择", 《地理信息世界》 * |
| 陶舒 等: "面向对象的高分辨率遥感影像数据的道路提取流程---地理国情普查数据的自动提取", 《遥感信息》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101840581B (en) | Method for extracting profile of building from satellite remote sensing image | |
| CN108985247B (en) | Multispectral image urban road identification method | |
| CN110309781B (en) | House damage remote sensing identification method based on multi-scale spectrum texture self-adaptive fusion | |
| CN107330875B (en) | Water body surrounding environment change detection method based on forward and reverse heterogeneity of remote sensing image | |
| CN110263717A (en) | It is a kind of incorporate streetscape image land used status determine method | |
| CN108596103A (en) | High resolution ratio satellite remote-sensing image building extracting method based on optimal spectrum Index selection | |
| CN112381013B (en) | Urban vegetation inversion method and system based on high-resolution remote sensing image | |
| Timilsina et al. | Mapping urban trees within cadastral parcels using an object-based convolutional neural network | |
| CN109063754A (en) | A kind of remote sensing image multiple features combining classification method based on OpenStreetMap | |
| CN109859219B (en) | High-resolution remote sensing image segmentation method combining phase and spectrum | |
| CN111626269A (en) | Practical large-space-range landslide extraction method | |
| CN114596495B (en) | Sentinel-2A remote sensing image-based landslide identification and automatic extraction method | |
| CN112017160A (en) | Multi-strategy combination-based multi-source remote sensing image road material fine extraction method | |
| CN109598202A (en) | A kind of object-based satellite image multi objective built-up areas extraction method | |
| Truax | Comparing spectral and object based approaches for classification and transportation feature extraction from high resolution multispectral imagery | |
| Lak et al. | A new method for road detection in urban areas using high-resolution satellite images and Lidar data based on fuzzy nearest-neighbor classification and optimal features | |
| CN114241321A (en) | Rapid and accurate identification method for high-resolution remote sensing image flat-topped building | |
| He et al. | Object-based distinction between building shadow and water in high-resolution imagery using fuzzy-rule classification and artificial bee colony optimization | |
| CN115205528A (en) | Feature selection method for geographic object-oriented image analysis | |
| CN110211128A (en) | Loess plateau terraced fields extracting method based on remote sensing image and DEM | |
| CN111882573B (en) | Cultivated land block extraction method and system based on high-resolution image data | |
| CN113159044A (en) | Deep learning-based road material identification method for convolutional neural network | |
| CN114120133A (en) | Urban road information extraction method for aggregating multiple factors | |
| CN115512159A (en) | Object-oriented high-resolution remote sensing image earth surface coverage classification method and system | |
| Fawzy et al. | AUTOMATIC INDICES BASED CLASSIFICATION METHOD FOR MAP UPDATING USING VHR 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 |