阅读说明：本技术 一种建筑物高度提取方法和装置 (Building height extraction method and device ) 是由 张晓东 胡腾云 石晓冬 李雪草 解鹏飞 孙道胜 何莲娜 吴兰若 于 2021-06-29 设计创作，主要内容包括：本发明提供了一种建筑物高度提取方法和装置,其中,该方法包括：通过获取目标建筑物的特征信息；其中,特征信息包括建筑面积、建筑周长、建筑面积周长比、后向散射系数、在可见光及近红外波段的反射率和平均海拔高度；将特征信息输入已训练的随机森林模型,得到目标建筑物的高度；其中,已训练的随机森林模型是根据建筑物样本的特征信息以及对应的高度标签训练得到的。本发明通过对多种与建筑相关的特征信息进行组合、测试、比较,确定最终输入随机森林模型进行建筑物高度提取的建筑物的特征信息,能够保证提取的建筑物的高度数据的精度,并且可以大批量地提取建筑物的高度数据。(The invention provides a building height extraction method and a building height extraction device, wherein the method comprises the following steps: obtaining characteristic information of a target building; the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude; inputting the characteristic information into the trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label. The invention determines the characteristic information of the building which is finally input into the random forest model for building height extraction by combining, testing and comparing various characteristic information related to the building, can ensure the accuracy of the extracted height data of the building, and can extract the height data of the building in large batch.)

1. A building height extraction method is characterized by comprising the following steps:

acquiring characteristic information of a target building; wherein the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude;

inputting the characteristic information into a trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

2. The building height extraction method according to claim 1, wherein obtaining the building area, the building perimeter, and the building area perimeter ratio of the target building comprises:

extracting contour information of the target building;

and acquiring the building area, the building perimeter and the building area perimeter ratio of the target building according to the contour information.

3. The building height extraction method according to claim 1, wherein obtaining the backscattering coefficient of the target building comprises:

acquiring polarized radar data of a C wave band of a sentinel No. 1 satellite corresponding to the target building, wherein the polarized radar data comprises backscattering coefficients in two polarization modes of vertical transmitting and vertical receiving and vertical transmitting and horizontal receiving;

and acquiring the backscattering coefficient of the target building according to the polarized radar data.

4. The building height extraction method according to claim 1, wherein obtaining the reflectivity of the target building in visible and near infrared bands comprises:

acquiring optical data of a sentinel No. 2 satellite corresponding to the target building;

and acquiring the reflectivity of the target building in visible light and near infrared bands according to the optical data.

5. The building height extraction method according to claim 1, wherein obtaining the average altitude of the target building comprises:

acquiring digital elevation model data of an advanced earth observation satellite corresponding to the target building;

and acquiring the average altitude of the target building according to the digital elevation model data.

6. The building height extraction method according to claim 1, further comprising:

dividing the building sample into a training set and a testing set;

inputting the characteristic information of the building sample in the training set into the random forest model, and adjusting the parameters of the random forest model according to the corresponding height label;

inputting the characteristic information of the building sample in the test set into the random forest model with the adjusted parameters, and testing the random forest model with the adjusted parameters according to the corresponding height label;

and determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

7. A building height extraction device, comprising:

the characteristic acquisition module is used for acquiring characteristic information of a target building; wherein the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude;

the height extraction module is used for inputting the characteristic information into a trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

8. The building height extraction apparatus according to claim 7, further comprising,

the model training module is used for dividing the building sample into a training set and a testing set;

inputting the characteristic information of the building sample in the training set into the random forest model, and adjusting the parameters of the random forest model according to the corresponding height label;

inputting the characteristic information of the building sample in the test set into the random forest model with the adjusted parameters, and testing the random forest model with the adjusted parameters according to the corresponding height label;

and determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

9. The building height extraction apparatus according to claim 7, wherein the feature acquisition module includes:

a first acquisition unit for extracting contour information of the target building;

the first processing unit is used for acquiring the building area, the building perimeter and the building area perimeter ratio of the target building according to the contour information; and/or the presence of a gas in the gas,

the second acquisition unit is used for acquiring polarized radar data of the C wave band of the sentinel No. 1 satellite corresponding to the target building, wherein the polarized radar data comprises backscattering coefficients in two polarization modes of vertical transmitting and vertical receiving and vertical transmitting and horizontal receiving;

and the second processing unit is used for acquiring the backscattering coefficient of the target building according to the polarized radar data.

10. The building height extraction apparatus according to claim 7, wherein the feature acquisition module includes:

the third acquisition unit is used for acquiring optical data of the sentinel No. 2 satellite corresponding to the target building;

the third processing unit is used for acquiring the reflectivity of the target building in visible light and near infrared bands according to the optical data; and/or the presence of a gas in the gas,

the fourth acquisition unit is used for acquiring digital elevation model data of an advanced earth observation satellite corresponding to the target building;

and the fourth processing unit is used for acquiring the average altitude of the target building according to the digital elevation model data.

Technical Field

The invention relates to the field of urban planning and construction, in particular to a building height extraction method and device.

Background

Due to the acceleration of the urbanization process and the continuous expansion of urban buildings, urban land resources are scarce, and cities are shifted to three-dimensional development. The research on the spatial form of the urban structure in the horizontal direction cannot meet the objective rule of urban development, and the research on the urban spatial structure needs to be changed from two-dimensional spatial layout to three-dimensional spatial development. As important basic data for building a three-dimensional digital city, building height information is important data for planning, construction project management, and various economic activities.

The method for acquiring the current building height comprises the following steps: the method comprises the steps of directly extracting the height of a building from an image, obtaining the height of the building from a building thematic Information database established by map data of an original two-dimensional Geographic Information System (GIS), estimating the height of the building from the number of floors of the building and the use property of the building, and the like.

Although the height of the building can be obtained by the above methods, the methods still have some defects. For example, the building height is directly extracted from the image, and the method is not suitable for automatic processing of mass data at present; the method comprises the steps of obtaining the height of a building from a building thematic information database established by map data of an original two-dimensional GIS, estimating the height of the building from the number of layers of the building and the use property of the building, and obtaining height data with lower accuracy.

Disclosure of Invention

The invention provides a building height extraction method, which is used for overcoming the defects that the height data of buildings cannot be extracted in a large batch and the accuracy of the obtained height data is low in the prior art, can realize the large-batch extraction of the height data of the buildings, and can ensure the extraction accuracy of the height data of the buildings.

In a first aspect, the present invention provides a building height extraction method, including: acquiring characteristic information of a target building; wherein the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude; inputting the characteristic information into a trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

According to the building height extraction method provided by the invention, the step of obtaining the building area, the building perimeter and the building area perimeter ratio of the target building comprises the following steps: extracting contour information of the target building; and acquiring the building area, the building perimeter and the building area perimeter ratio of the target building according to the contour information.

According to the building height extraction method provided by the invention, the step of obtaining the backscattering coefficient of the target building comprises the following steps: acquiring polarized radar data of a C wave band of a sentinel No. 1 satellite corresponding to the target building, wherein the polarized radar data comprises backscattering coefficients in two polarization modes of vertical transmitting and vertical receiving and vertical transmitting and horizontal receiving; and acquiring the backscattering coefficient of the target building according to the polarized radar data.

According to the building height extraction method provided by the invention, the method for acquiring the reflectivity of the target building in visible light and near infrared bands comprises the following steps: acquiring optical data of a sentinel No. 2 satellite corresponding to the target building; and acquiring the reflectivity of the target building in visible light and near infrared bands according to the optical data.

According to the building height extraction method provided by the invention, the step of obtaining the average altitude of the target building comprises the following steps: acquiring digital elevation model data of an advanced earth observation satellite corresponding to the target building; and acquiring the average altitude of the target building according to the digital elevation model data.

The building height extraction method provided by the invention further comprises the following steps: dividing the building sample into a training set and a testing set; inputting the characteristic information of the building sample in the training set into the random forest model, and adjusting the parameters of the random forest model according to the corresponding height label; inputting the characteristic information of the building sample in the test set into the random forest model with the adjusted parameters, and testing the random forest model with the adjusted parameters according to the corresponding height label; and determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

In a second aspect, the present invention provides a building height extraction apparatus comprising: the characteristic acquisition module is used for acquiring characteristic information of a target building; wherein the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude; the height extraction module is used for inputting the characteristic information into a trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

The building height extraction device further comprises a model training module, a building height extraction module and a building height extraction module, wherein the model training module is used for dividing the building sample into a training set and a testing set; inputting the characteristic information of the building sample in the training set into the random forest model, and adjusting the parameters of the random forest model according to the corresponding height label; inputting the characteristic information of the building sample in the test set into the random forest model with the adjusted parameters, and testing the random forest model with the adjusted parameters according to the corresponding height label; and determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

According to the building height extraction apparatus provided by the present invention, the feature acquisition module includes: a first acquisition unit for extracting contour information of the target building; the first processing unit is used for acquiring the building area, the building perimeter and the building area perimeter ratio of the target building according to the contour information; and/or the second acquisition unit is used for acquiring polarized radar data of the C wave band of the sentinel No. 1 satellite corresponding to the target building, wherein the polarized radar data comprises backscattering coefficients in two polarization modes of vertical transmitting and vertical receiving and vertical transmitting and horizontal receiving; and the second processing unit is used for acquiring the backscattering coefficient of the target building according to the polarized radar data.

According to the building height extraction apparatus provided by the present invention, the feature acquisition module further includes: the third acquisition unit is used for acquiring optical data of the sentinel No. 2 satellite corresponding to the target building; the third processing unit is used for acquiring the reflectivity of the target building in visible light and near infrared bands according to the optical data; and/or a fourth acquisition unit, configured to acquire digital elevation model data of an advanced earth observation satellite corresponding to the target building; and the fourth processing unit is used for acquiring the average altitude of the target building according to the digital elevation model data.

The invention provides a building height extraction method and a building height extraction device, which are used for extracting the height of a building by acquiring the characteristic information of a target building; the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude; inputting the characteristic information into the trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label. The invention determines the characteristic information of the building which is finally input into the random forest model for building height extraction by combining, testing and comparing various characteristic information related to the building, and can ensure the accuracy of the extracted height data of the building; the random forest model can process data with very high dimensionality, is suitable for parallel computation, and can extract height data of buildings in large batch; and the feature information of the building is input into the trained random forest model to obtain the height data of the building, and the operation is simple and convenient.

Drawings

In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic flow diagram of a building height extraction method provided by the present invention;

FIG. 2 is a schematic diagram of a random forest model generation process provided by the present invention;

FIG. 3 is a schematic flow chart of a random forest model training process provided by the present invention;

FIG. 4 is a schematic diagram of an application scenario for training a random forest model according to the present invention;

fig. 5 is a schematic structural view of a building height extraction apparatus provided by the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In the present invention, a building may be a single building or a group of buildings.

Fig. 1 is a schematic flow chart of a building height extraction method provided by the invention. As shown in fig. 1, the building height extraction method includes the following steps:

s101, acquiring characteristic information of a target building; the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude;

s102, inputting the characteristic information into the trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

In step S101, the building area, the building perimeter, and the building area perimeter ratio may be obtained based on the contour feature of the target building, for example, the contour feature of the target building may be extracted through edge detection, or the contour feature of the target building may be extracted through a contour tracking algorithm. The method for acquiring the contour features of the target building is not limited in the embodiment of the invention.

The backscattering coefficient represents the ratio of the backscattering cross-section to the incident light cross-section. The backscattering coefficient of high-rise building areas, such as central business areas, is obviously higher than that of low-rise building areas, such as residential areas, and the building height has a clear correlation with the backscattering coefficient. The polarized radar data includes backscatter coefficients in two polarization modes of vertical transmission vertical reception (VV) and vertical transmission horizontal reception (VH), and thus the backscatter coefficients may be obtained based on polarized radar data of the C band, which may be acquired by a sentinel one satellite. The sentinel satellite No. 1 is an earth observation satellite in the Colkini program of the European space agency, consists of two satellites, carries a C-band synthetic aperture radar and can provide continuous images.

When a light source irradiates the surface of an object, the object can selectively reflect electromagnetic waves with different wavelengths, and the reflectivity is the ratio of luminous flux reflected by the object to luminous flux incident on the object in a certain waveband and is an essential attribute of the surface of the object. The embodiment of the present invention does not limit the selection of the visible light. Red, green and blue light can be generally selected as visible light, because buildings can generate higher reflectivity for red, green and blue bands, so that reflectivity data can be easily acquired. The reflectivity of the target building in the visible and near infrared bands can be obtained based on the optical data of the sentinel satellite number 2. The optical data of the sentinel 2 satellite is multispectral remote sensing data with the highest spatial resolution obtained by the sentinel 2 satellite, and the data is high in spatial resolution and spectral resolution. The sentinel 2 satellite is a high-resolution multispectral imaging satellite, carries a multispectral imager, is used for land monitoring, can provide images of vegetation, soil and water coverage, inland waterway, coastal areas and the like, and can also be used for emergency rescue service.

The average altitude of the target building may be obtained based on digital elevation model data of advanced earth observation satellites. The digital elevation model is used for realizing digital simulation of a terrain curved surface or digital representation of terrain surface morphology through limited terrain elevation. The digital elevation model is high-precision global digital earth surface model data freely provided by the Japanese space aviation research and development organization, the horizontal resolution of the digital elevation model is 30 meters, the elevation precision is 5 meters, the digital elevation model is the most accurate 3D map in the world at present, and all global land scales are covered.

In step S102, the random forest model is a prediction model. FIG. 2 is a schematic diagram of a random forest model generation process provided by the present invention. As shown in fig. 2, the random forest model generation process is: performing replaced random sampling on an original sample set for N times by a Bootstrap aggregation (Bagging) sampling method, wherein the value of N is more than 2, and obtaining N sub-sample sets with the size of N; taking each subsample set as a training sample of a decision tree, taking the training sample as a root node of the corresponding decision tree, and then executing a top-down greedy search algorithm from the root node to obtain n trained decision trees; and combining the n trained decision trees to obtain a random forest model. Because each decision tree in the random forest model selects part of samples and part of characteristics, overfitting can be avoided to a certain extent, and the accuracy of prediction is improved; moreover, samples are randomly selected and characteristics are randomly selected in each decision tree, so that the noise immunity is good, and the performance is stable; because a plurality of decision trees are generated by a plurality of sample subsets, the random forest model can process data with high dimensionality, does not need to make feature selection and is suitable for parallel computation.

The building height extraction method provided by the invention comprises the steps of obtaining characteristic information of a target building; the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude; inputting the characteristic information into the trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label. The invention determines the characteristic information of the building which is finally input into the random forest model for building height extraction by combining, testing and comparing various characteristic information related to the building, and can ensure the accuracy of the extracted height data of the building; the random forest model can process data with very high dimensionality, is suitable for parallel computation, and can extract height data of buildings in large batch; and the feature information of the building is input into the trained random forest model to obtain the height data of the building, and the operation is simple and convenient.

FIG. 3 is a flow chart of a random forest model training process provided by the present invention. As shown in fig. 3, the training process of the random forest model includes:

step S301, dividing the building sample into a training set and a test set building;

step S302, inputting the characteristic information of the building sample in the training set into a random forest model, and adjusting the parameters of the random forest model according to the corresponding height label;

step S303, inputting the characteristic information of the building sample in the test set into the random forest model after parameter adjustment, and testing the random forest model after parameter adjustment according to the corresponding height label;

and S304, determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

In step S301, the building sample includes characteristic information and a corresponding height label. The height label may be height information of the building obtained from information disclosed on the network, or may be height information of the building obtained from an archive or a planning office of a city where the building is located. The embodiment of the invention does not limit the method for obtaining the height label. The building sample can be divided into a test set and a training set according to a certain proportion, and the division proportion is not limited by the embodiment of the invention. This division ratio is typically taken to be 7: 3.

In step S302, inputting the characteristic information of the building sample in the training set into a random forest model, outputting a height data by the random forest model, comparing the output height data with a corresponding height label, and outputting the height data obtained by the random forest model if the condition is met; and if the condition is not met, adjusting the parameters of the random forest model. Embodiments of the present invention are not limited in the manner in which the outputted height data is compared to the height label. For example, the difference between the two may be compared with a preset value, or the variance between the two may be compared with a preset value.

In step S303, the random forest model with the adjusted parameters is tested, and the testing process is similar to the training process, so that the description is not repeated.

In step S304, it is determined whether the random forest model after parameter adjustment is a trained random forest model according to the accuracy in the test result. The embodiment of the invention does not limit the accuracy in the test result. For example, the ratio may be taken as 100%, or the ratio may be taken as 70%. And if the accuracy in the test result is determined to be more than 70%, judging that the random forest model after the parameters are adjusted is the trained random forest model.

When the feature information of the building is selected, the feature information of the finally selected building can be determined by performing combined test on various feature information related to the building and comparing test results.

When a plurality of kinds of characteristic information related to buildings are tested in a combined mode, backscattering coefficients in two polarization modes of vertical transmission and vertical reception (VV) and vertical transmission and horizontal reception (VH) and reflectivities of visible light and near infrared wave bands are obtained according to reflection of electromagnetic waves and form a group of data. The maximum value, the minimum value and the average value can be selected as the characteristic data of the backscattering coefficient and the reflectivity of visible light and near infrared bands for testing; or only the average value can be selected as the characteristic data of the backscattering coefficient and the reflectivity of the visible light and the near infrared band for testing; alternatively, the maximum value, the minimum value, the average value and other values can be selected as characteristic data of the backscattering coefficient and the reflectance of the visible light and near infrared band, for example, the maximum value, the minimum value, the average value, the maximum value of 5%, the maximum value of 25%, the maximum value of 50%, the maximum value of 75% and the maximum value of 95%, and the test can be performed.

When a plurality of kinds of characteristic information related to buildings are combined and tested, the characteristic information of the perimeter ratio of the building area, the reflectivity of yellow, red, blue and near infrared bands in visible light can be combined; or combining characteristic information of the building area perimeter ratio, the backscattering coefficients in two polarization modes of vertical emission vertical receiving (VV) and vertical emission horizontal receiving (VH), and the reflectivity of yellow-red blue light and near-infrared band in visible light; or the characteristic information of the building area perimeter ratio, the backscattering coefficients in two polarization modes of vertical transmission and vertical reception (VV) and vertical transmission and horizontal reception (VH), and the reflectivity of blue-green-red light and near infrared bands in visible light can be combined.

When the characteristic information of the backscattering coefficient, the reflectivity of blue-green red light and near infrared wave bands in visible light and the average altitude in two polarization modes of the building area perimeter ratio, the vertical transmitting vertical receiving (VV) and the vertical transmitting horizontal receiving (VH) are combined, wherein the backscattering coefficient, the reflectivity of the visible light and the near infrared wave bands are selected, the maximum value, the minimum value and the average value are selected, the trained random forest model is input for testing, and the test result is shown in table 1. Wherein, 1-3, 4-6, 7-9 and 10-represent the number of floors, the first row represents the actual value, and the first column represents the predicted value. For example, the numeral 37300 in the second row and the second column means that the number of test samples with the predicted height and the actual height of 1-3 layers is 37300. The number 42346 in the second row and the sixth column means that the total number of test samples predicted at layers 1 to 3 is 42346, and the seventh column indicates the accuracy. The numeral 88.08% in the seventh column of the second row indicates that the ratio of the predicted value and the actual value in each of the 1-3 layers to the total number of test samples in the predicted 1-3 layers is 88.08%. The seventh row represents recall, which is a measure of coverage, how many actual positive examples are predicted as positive examples. The numeral 87.33% in the second column of the seventh row indicates the ratio of the predicted value and the actual value of the layers 1-3 to the total number of test samples of the actual value of the layers 1-3. The accuracy of all predicted values statistically in table 1 was 76.32%.

Table 1 is a first test data sheet

	1-3	4-6	7-9	10～	Sum	Precision
							1-3	37300	3822	713	511	42346	88.08％
4-6	4801	5284	581	260	10926	48.36％
							7-9	275	653	289	187	1404	20.58％
10～	338	881	997	2313	4529	51.07％
							Sum	42714	10640	2580	3271	59205
Recall	87.33％	49.66％	11.20％	70.71％
							Accuracy(OA)	76.32％
Kappa1	0.469

When the characteristic information of the backscattering coefficient, the reflectivity of blue-green red light and near infrared wave bands in visible light and the average altitude in two polarization modes of building area, building perimeter, building area perimeter ratio, vertical transmission vertical reception (VV) and vertical transmission horizontal reception (VH) are combined, wherein the backscattering coefficient, the reflectivity of visible light and near infrared wave bands are selected, the maximum value, the minimum value, the average value, the maximum value of 5%, the maximum value of 25%, the maximum value of 50%, the maximum value of 75% and the maximum value of 95% are selected, and the training random forest model is input for testing, and the test result is shown in Table 2. The columns and rows in Table 2 have the same meanings as in Table 1 and are not described in detail. The accuracy of all predicted values statistically obtained in table 2 was 79.46%, which was improved compared to table 1.

Table 2 is a second test data sheet

	1-3	4-6	7-9	10～	Sum	Precision
							1-3	56599	3448	253	434	60734	93.19％
4-6	5868	7236	992	1434	15530	46.59％
							7-9	1137	1284	509	1319	4249	11.98％
10～	448	369	205	2169	3191	67.97％
							Sum	64052	12337	1959	5356	83704
Recall	88.36％	58.65％	25.98％	40.50％
							Accuracy(OA)	79.46％
Kappa1	0.504

The method comprises the steps of performing combined test on various feature information related to buildings, comparing test results, combining the feature information with the highest accuracy, namely combining the feature information with the highest accuracy, namely building area, building perimeter, building area perimeter ratio, backward scattering coefficients in two polarization modes of vertical emission vertical reception (VV) and vertical emission horizontal reception (VH), reflectivity of blue-green red light and near infrared bands in visible light and average altitude, and determining the feature information for inputting a trained random forest model to perform building height extraction.

Fig. 4 is a schematic diagram of an application scenario of training a random forest model according to the present invention. As shown in fig. 4, the application scenario includes the following steps: step 1, collecting building outlines and corresponding height data of buildings in typical areas of Beijing city; step 2, obtaining the ratio of the building area, the building perimeter and the building area perimeter according to the building outline; step 3, obtaining a backscattering coefficient of the building in the typical area of Beijing City according to the collected polarization radar data of the building in the typical area of Beijing City; step 4, acquiring the reflectivity of the buildings in the typical area of Beijing City in red, green and near infrared bands according to the collected optical data of the buildings in the typical area of Beijing City; step 5, acquiring the average altitude of the buildings in the typical area of Beijing City according to the collected digital elevation model data of the buildings in the typical area of Beijing City; and 6, inputting the building area, the building perimeter, the building area perimeter ratio, the backscattering coefficient, the reflectivity in red, green, blue and near infrared bands, the average altitude and the corresponding height data of the building in the typical area of Beijing city obtained in the steps 1-5 into the constructed random forest model, and training the random forest model to obtain the trained random forest model.

Fig. 5 is a schematic view of a building height extraction apparatus provided by the present invention. As shown in fig. 5, the building height extracting apparatus includes:

a characteristic obtaining module 501, configured to obtain characteristic information of a target building; the characteristic information comprises building area, building perimeter, building area perimeter ratio, backscattering coefficient, reflectivity in visible light and near infrared wave bands and average altitude;

a height extraction module 502, configured to input the feature information into the trained random forest model to obtain the height of the target building; and the trained random forest model is obtained by training according to the feature information of the building sample and the corresponding height label.

Optionally, the feature obtaining module 501 includes:

a first acquisition unit for extracting contour data of a target building;

and the first processing unit is used for acquiring the building area, the building perimeter and the building area perimeter ratio of the target building according to the contour information.

Optionally, the feature obtaining module 501 further includes:

the second acquisition unit is used for acquiring polarized radar data of the C wave band of the sentinel No. 1 satellite corresponding to the target building, wherein the polarized radar data comprises backscattering coefficients in two polarization modes of vertical transmitting and vertical receiving and vertical transmitting and horizontal receiving;

and the second processing unit is used for acquiring the backscattering coefficient of the target building according to the polarized radar data.

Optionally, the feature obtaining module 501 further includes:

the third acquisition unit is used for acquiring optical data of the sentinel No. 2 satellite corresponding to the target building;

and the third processing unit is used for acquiring the reflectivity of the target building in visible light and near infrared bands according to the optical data.

Optionally, the feature obtaining module 501 further includes:

the fourth acquisition unit is used for acquiring digital elevation model data of an advanced earth observation satellite corresponding to the target building;

and the fourth processing unit is used for acquiring the average altitude of the target building according to the digital elevation model data.

Optionally, the building height extraction device, further comprises,

the model training module is used for dividing the building sample into a training set and a testing set; inputting the characteristic information of the building sample in the training set into a random forest model, and adjusting the parameters of the random forest model according to the corresponding height label; inputting the characteristic information of the building sample in the test set into the random forest model after the parameters are adjusted, and testing the random forest model after the parameters are adjusted according to the corresponding height label; and determining the trained random forest model according to the test result of the random forest model after the parameters are adjusted on the test set.

The above-described embodiments of the apparatus are merely illustrative, and the modules described as separate parts may or may not be physically separate, and the parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods of the various embodiments or some parts of the embodiments.

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.