Urban night light environment remote sensing monitoring method combining unmanned aerial vehicle and SDGSAT-1 satellite data

Abstract

The invention provides an urban night light environment remote sensing monitoring method combining unmanned aerial vehicle and SDGSAT-1 satellite data, which comprises the following steps of: estimating typical small-region high-resolution night ground surface incident illumination based on an unmanned aerial vehicle night multispectral image, and then upscaling the high-resolution ground surface illumination estimated by the unmanned aerial vehicle to the resolution of an SDGSAT-1 satellite image; and combining with SDGSAT-1 satellite data to train a surface illumination model, estimating the night surface illumination of the whole city, and monitoring and evaluating the night light environment of the city. According to the invention, the urban light environment condition can be effectively monitored, and a scientific basis is provided for urban planning and light environment governance.

Description

Technical Field

[0001] The invention relates to the technical field of nighttime light environment, and in particular to a remote sensing monitoring method for urban nighttime light environment combining unmanned aerial vehicle (UAV) and SDGSAT-1 satellite data. Background Art

[0002] Artificial light sources are widely used in modern society, bringing many conveniences to human life. Nighttime lighting can improve road visibility and facilitate people's travel at night. Businesses use nighttime lights to create a good consumption environment, which not only enriches residents' nightlife, but also further stimulates consumption and promotes urban economic development. In some areas, insufficient lighting will affect residents' nightlife. On the other hand, excessive artificial lighting can also cause light pollution. In order to create a scientific and reasonable urban nighttime light environment, it is necessary to monitor and control urban nighttime lighting.

[0003] Traditional nighttime light environment monitoring methods usually rely on illuminometers and other equipment for field fixed-point observation or mobile observation. This type of ground monitoring method can only obtain limited point-scale observation data. However, the distribution of light environment has significant spatial heterogeneity, and limited measurement samples cannot effectively reflect the spatial distribution characteristics of nighttime light environment at the urban scale. Satellite remote sensing can provide large-scale and long-term spatial continuous observation. At present, nighttime light remote sensing satellites such as DMSP / OLS, NPP / VIIRS, Luojia 1-01, and JL1-3B have been put into use. However, the resolution of most nighttime light remote sensing data is low, such as DMSP / OLS with a spatial resolution of 2.8 kilometers, NPP / VIIRS with a spatial resolution of 750 meters, and Luojia 1-01 with a spatial resolution of 130 meters, which cannot effectively reflect the spatial details of the light environment. Although JL1-3B has a resolution of up to 0.92 meters, its width is only 11km, which cannot meet the observation needs at the urban scale. Most nighttime light remote sensing sensors also have only one channel, which cannot fully reflect the spectral characteristics of different types of light sources. The newly launched SDGSAT-1 satellite low-light band can provide one 10-meter resolution panchromatic band and three 40-meter resolution multispectral images, and the imaging width is up to 300 kilometers, which has the potential for fine-grained monitoring of urban-scale night light environment monitoring. In addition, due to the spatial heterogeneity of the night light environment, the spatial scale difference between satellite remote sensing image pixels and single-point observations will bring great uncertainty to the estimation. The emergence of drones provides a new technical means for monitoring the night light environment. Drones can not only provide high-resolution night light images, but also have relatively low costs and are flexible to use. However, drones have limited range and can only carry out small-scale observations, which is not enough to cover the entire city. Summary of the invention

[0004] The purpose of the present invention is to provide a remote sensing monitoring method for urban nighttime light environment by combining unmanned aerial vehicle and SDGSAT-1 satellite data, which can effectively monitor the urban nighttime light environment and provide a scientific basis for urban planning and light environment management.

[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is: A method for remote sensing monitoring of urban nighttime light environment using combined unmanned aerial vehicle (UAV) and SDGSAT-1 satellite data, the method comprising the following steps: Select an observation area, conduct drone aerial photography and field observation, obtain multispectral remote sensing images at night and during the day, and the observed illumination values ​​of the observation points; calibrate the multispectral remote sensing images during the day to obtain the surface reflectance information of the observation area; Combined with the surface reflectivity information of the observation area, the nighttime surface reflected radiance received by the drone is converted into surface incident radiance; The surface incident radiance of the observation point is taken as the independent variable, and the observed illuminance value of the observation point is taken as the dependent variable. The random forest algorithm is used to train the first surface incident illuminance estimation model based on UAV night images. The trained first surface incident illuminance estimation model is applied to the surface incident radiance corresponding to the UAV to obtain the UAV surface incident illuminance of the observation area, and it is upscaled to the same resolution as the SDGSAT-1 satellite remote sensing image. The SDGSAT-1 satellite surface incident radiance of the observation area is calculated based on the SDGSAT-1 satellite remote sensing images after radiation calibration, surface reflectance data, building coverage ratio and tree coverage ratio; The SDGSAT-1 satellite surface incident radiance in the observation area is taken as the independent variable, and the corresponding upscaled UAV surface incident illuminance is taken as the dependent variable. The random forest algorithm is used to train the second surface incident illuminance estimation model based on the SDGSAT-1 satellite. The trained second surface incident illuminance estimation model is applied to the surface incident radiance of the SDGSAT-1 satellite in the entire city, and the night-time incident surface illuminance distribution map of the entire city is obtained.

[0006] Furthermore, the process of conducting drone aerial observation and field observation includes the following steps: While the SDGSAT-1 satellite is passing by at night, an observation area is selected and a multispectral imager is used to conduct flight observations to obtain nighttime multispectral remote sensing images.

[0007] During the UAV flight observation, a handheld illuminance meter was used to measure and record the downlink illuminance value of the observation point horizontally upward, and a handheld RTK device was used to record the latitude and longitude position data of the observation point; Use a multispectral imager mounted on a drone to conduct aerial observations in the same observation area during the daytime at a similar time phase, obtain daytime multispectral remote sensing images in the observation area, and take calibration whiteboard images before and after the flight observation; The multispectral remote sensing images at night and during the day are stitched together using UAV orthophoto processing software.

[0008] Furthermore, combined with the surface reflectivity information of the observation area, the process of converting the surface reflected radiance received by the drone into the surface incident radiance includes the following steps: The data of the three visible light bands of RGB in the multispectral remote sensing images of the drone at night are retained to correct the multispectral remote sensing images of the drone at night; Convert the surface reflected radiance received by the drone into the surface incident radiance: ; In the formula, is the surface incident radiance of the UAV in the i-th band; is the radiation brightness at the entrance pupil of the drone in the i-th band; is the surface reflectivity of the UAV in the i-th band.

[0009] Furthermore, the process of using the random forest algorithm to train the first surface incident illumination estimation model based on UAV nighttime images includes: According to the latitude and longitude information of the measured points, the surface incident radiance values ​​of the three visible light bands of RGB in the corresponding multispectral remote sensing images of the UAV at night are extracted; Taking the surface incident radiance of the observation point as the independent variable and the observed illuminance value of the observation point as the dependent variable, the random forest algorithm is used to train the first surface incident illuminance estimation model based on UAV night images. During the training process, grid search is used to optimize the model parameters, and ten-fold cross validation is used to verify the model accuracy.

[0010] Furthermore, the process of upscaling the UAV surface incident illumination in the observation area to the same resolution as the SDGSAT-1 satellite remote sensing image includes the following steps: The 10-meter grids are cycled one by one, and the average surface incident illumination of non-building and tree pixels in each 10-meter grid is counted to obtain the surface incident illumination map of the observation area with a 10-meter resolution.

[0011] Furthermore, remote sensing processing software is used to pre-process the SDGSAT-1 satellite images by mosaicking and cropping them, and radiometric calibration is performed on them to convert the grayscale values ​​into radiance values: ; In the formula, is the radiance of the satellite at the entrance pupil in the i-th band; is the gray value of the satellite's i-th band; and are the satellite i-th band gain and offset respectively.

[0012] Furthermore, the process of calculating the SDGSAT-1 satellite surface incident radiance of the observation area includes the following steps: The three visible light bands of RGB with a resolution of 40m and the panchromatic band with a resolution of 10m in the SDGSAT-1 satellite were fused by Gram-Schmidt orthogonal transformation to obtain the three visible light band images with a resolution of 10m. The surface reflectance data of the three visible light bands of RGB are extracted from the satellite daytime surface reflectance products of the adjacent phase; the building vector data and tree cover data are superimposed on the SDGSAT-1 pixels, and the building coverage and tree cover in each 10-meter resolution pixel are calculated; The SDGSAT-1 satellite surface incident radiance in the observation area is calculated using the following formula: ; In the formula, is the surface incident radiation brightness of the satellite in the i-th band; represents the radiance of the satellite at the entrance pupil in the i-th band; is the surface reflectivity of the satellite’s i-th band; is the building coverage; is the tree cover.

[0013] Furthermore, the method further comprises: The nighttime surface incident illumination distribution map obtained based on remote sensing represents the urban light environment. The nighttime insufficiently and excessively illuminated areas in the city are identified by combining the surface incident illumination distribution map with the city map and functional zoning.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention combines the remote sensing monitoring method of urban night light environment with unmanned aerial vehicle and SDGSAT-1 satellite data, using unmanned aerial vehicle images as the intermediary, effectively reduces the uncertainty caused by the spatial scale difference between ground observation and satellite image pixels; and uses building coverage, tree coverage and surface reflectivity to correct the fused SDGSAT-1 multi-spectral channel, so as to accurately estimate the surface incident radiance, and effectively monitor the urban night light environment. The method provided by the present invention can accurately reflect the spatial distribution of the urban night light environment, and provide technical support for the rational control of urban night lighting. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1A flow chart of a method for remote sensing monitoring of urban nighttime light environment using combined unmanned aerial vehicle and SDGSAT-1 satellite data of the present invention; Figure 2 Scatter plot between the estimated and observed values ​​of surface incident illumination based on drone data; Figure 3 The spatial distribution diagram of the nighttime surface illumination in two observation areas, where a) corresponds to typical area 1 and b) corresponds to typical area 2; Figure 4 It is a scatter plot between the estimated and measured values ​​of surface incident illumination based on SDGSAT-1 satellite data; Figure 5 This is the spatial distribution map of surface illumination at night in Nanjing; Figure 6 This is the radiation calibration coefficient diagram of the SDGSAT-1 satellite low-light band. DETAILED DESCRIPTION

[0016] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0017] The present invention discloses a method for remote sensing monitoring of urban nighttime light environment by combining unmanned aerial vehicles and SDGSAT-1 satellite data, the method comprising the following steps: Select an observation area, conduct drone aerial photography and field observation, obtain multispectral remote sensing images at night and during the day, and the observed illumination values ​​of the observation points; calibrate the multispectral remote sensing images during the day to obtain the surface reflectance information of the observation area; Combined with the surface reflectivity information of the observation area, the surface reflected radiance received by the drone is converted into the surface incident radiance; The surface incident radiance of the observation point is taken as the independent variable, and the observed illuminance value of the observation point is taken as the dependent variable. The random forest algorithm is used to train the first surface incident illuminance estimation model based on UAV night images. The trained first surface incident illuminance estimation model is applied to the surface incident radiance corresponding to the UAV to obtain the UAV surface incident illuminance of the observation area, and it is upscaled to the same resolution as the SDGSAT-1 satellite remote sensing image. The SDGSAT-1 satellite surface incident radiance of the observation area is calculated based on the SDGSAT-1 satellite remote sensing images after radiation calibration, surface reflectance data, building coverage ratio and tree coverage ratio; The SDGSAT-1 satellite surface incident radiance in the observation area is taken as the independent variable, and the corresponding upscaled UAV surface incident illuminance is taken as the dependent variable. The random forest algorithm is used to train the second surface incident illuminance estimation model based on the SDGSAT-1 satellite. The trained second surface incident illuminance estimation model is applied to the surface incident radiance of the SDGSAT-1 satellite in the entire city, and the night-time incident surface illuminance distribution map of the entire city is obtained.

[0018] The present invention first estimates the high-resolution nighttime surface illumination of a typical small area based on the nighttime multispectral imagery of an unmanned aerial vehicle, then upscales it to the resolution of SDGSAT-1 satellite data to train a nighttime surface illumination estimation model based on SDGSAT-1 satellite data, estimates the nighttime surface incident illumination of the entire city, and monitors and evaluates the nighttime light environment conditions of the city.

[0019] 1) UAV and ground observation While the SDGSAT-1 satellite is passing by at night, a typical small area is selected for drone aerial observation. The multispectral imager is used to carry out flight observation, and the flight area, route, altitude, heading and lateral overlap, time interval and other parameters are set through ground station software such as DJI GS Pro to obtain nighttime multispectral remote sensing images.

[0020] During the UAV flight observation, the ground observation experiment was carried out simultaneously. The downlink illumination value of the observation point was measured and recorded horizontally using a handheld illuminance meter, and the latitude and longitude position data of the observation point was recorded using a handheld RTK device.

[0021] During the daytime at a similar time, a drone equipped with a multispectral instrument was used to conduct aerial observations in the same observation area to obtain daytime multispectral remote sensing images in a typical area. Calibration whiteboard images were also taken before and after the flight observation.

[0022] 2) Data processing The drone orthophoto processing software was used to stitch the multispectral remote sensing images at night and during the day. The multispectral remote sensing images during the day were calibrated based on the whiteboard information to obtain the surface reflectance information of the observation area. The nighttime lighting environment only considers the visible light range, so only the data of the three visible light bands of the multispectral image are retained.

[0023] The SDGSAT-1 satellite images were pre-processed by mosaicking and cropping using remote sensing processing software, and radiometric calibration was performed to convert the grayscale values ​​into radiance values: (1); In the formula, is the radiance of the satellite at the entrance pupil in the i-th band; is the gray value of the satellite's i-th band; and are the gain and offset of the satellite band i respectively. The SDGSAT-1 satellite low-light band radiation calibration coefficient is as follows: Figure 6 shown.

[0024] 3) Estimation of surface incident illumination in typical areas based on drones The radiation signal received by the drone is the light reflected from the ground, and the night light environment mainly depends on the incident light rather than the reflected light. Considering the high resolution of drone images and the absence of mixed pixel effects, the surface reflected radiation brightness received by the drone is converted into the surface incident radiation brightness.

[0025] (2); In the formula, is the surface incident radiance of the UAV in the i-th band; is the radiation brightness at the entrance pupil of the drone in the i-th band; is the surface reflectivity of the UAV in the i-th band.

[0026] According to the latitude and longitude information of the measured points, the surface incident radiance values ​​of the three visible light bands after the corresponding drone night images were corrected were extracted. With the surface incident radiance as the independent variable and the field observed illuminance value as the dependent variable, the random forest algorithm was used to train the surface incident illuminance estimation model based on drone night images. During the training process, the model parameters were optimized using grid search, and the model accuracy was verified using ten-fold cross validation. The optimal estimation model was determined and applied to the drone surface incident radiance data to obtain the surface incident illuminance map of a typical small area.

[0027] 4) Surface illumination estimation of the study area based on SDGSAT-1 satellite There is a large spatial scale difference between the multispectral pixels of the 40m resolution of the SDGSAT-1 satellite and the ground observation points. If the estimation is performed directly, it will bring great uncertainty to the result. Therefore, the three multispectral bands of the 40m resolution of the SDGSAT-1 satellite and the panchromatic band of the 10m resolution are firstly fused by Gram-Schmidt orthogonal transformation to obtain the three multispectral band images with a resolution of 10m; secondly, the high-resolution surface incident illumination of the typical area estimated by the UAV is upscaled to 10m resolution, and then combined with the SDGSAT-1 satellite data of the corresponding area to train the illumination estimation model based on the SDGSAT-1 satellite images, so as to better eliminate the deviation of the results.

[0028] Considering that the 10-meter resolution pixel of the SDGSAT-1 satellite may contain multiple types of ground objects such as the ground, buildings, and trees, the signal received by the satellite sensor is a comprehensive signal of the light signals of various objects within the 10-meter pixel. However, the satellite receives reflected light from the ground, there is almost no light on the top of the building, and the tree canopy will block the reflected light. These factors will cause the satellite received signal to deviate from the actual light environment, and the influence of buildings and tree canopies needs to be removed.

[0029] The remote sensing sensor receives the incident light intensity reflected from the ground surface, while the buildings and tree-covered areas in the pixel can be considered to be without light. Therefore, the apparent radiance can be expressed as: (3); In the formula, is the radiance of the satellite at the entrance pupil in the i-th band; is the surface incident radiation brightness of the satellite in the i-th band; is the surface reflectivity of the satellite’s i-th band; is the building coverage; is the tree cover.

[0030] By transforming formula (3), we can obtain the modified formula of surface incident radiance: (4); The surface reflectance data of three visible light bands were extracted from the satellite daytime surface reflectance product of the adjacent phase; the building vector data and tree cover data were superimposed on the SDGSAT-1 pixels, and the building coverage and tree cover in each 10-meter resolution pixel were calculated. The corrected SDGSAT-1 surface incident radiance was calculated using formula (4).

[0031] The high-resolution surface incident illumination of a typical small area estimated by UAV remote sensing is scaled to 10-meter resolution, which is consistent with SDGSAT-1 data. In the upscaling process, the 10-meter grid is cycled, and the average surface incident illumination of non-buildings and trees in each 10-meter grid is counted to obtain a surface incident illumination map of a typical small area with a resolution of 10 meters.

[0032] The surface incident radiance of the three bands of the corrected SDGSAT-1 satellite in a typical small area was extracted as the independent variable, and the corresponding 10-meter resolution UAV surface incident illuminance was used as the dependent variable. The surface incident illuminance estimation model was trained using the random forest algorithm. During the training process, the model parameters were optimized using grid search, and the model accuracy was verified using ten-fold cross validation. The optimal estimation model was determined and applied to the corrected SDGSAT-1 satellite surface incident radiance of the entire city, and the nighttime incident surface illuminance distribution map of the entire city was obtained.

[0033] 5) Night light environment analysis The nighttime surface incident illuminance distribution map obtained based on remote sensing represents the urban light environment. The insufficiently and excessively illuminated areas in the city at night are identified by combining the surface incident illuminance distribution map with the city map and functional zoning, thereby effectively evaluating the city's light environment level. This provides an important decision-making basis for the subsequent scientific planning and management of nighttime lighting.

[0034] Examples The present invention combines drone observation and SDGSAT-1 satellite remote sensing data to estimate the surface incident illumination and monitor the urban night light environment. The following are specific implementation steps for examples. Figure 1 .

[0035] 1) Two typical small areas (typical area 1 and typical area 2) were selected, and drone observations and field observations were carried out simultaneously during the nighttime transit imaging of the SDGSAT-1 satellite. The nighttime multispectral image data of the study area was obtained using the DJI Phantom 4 multispectral drone. The drone flight altitude was set to 400m, and the heading overlap rate and side overlap rate were set to 75%. During the drone aerial imaging process, ground observations were carried out simultaneously. The incident illumination value of each observation point was measured upward using a TES-1399R handheld illuminance meter, and the latitude and longitude information of the measured points were recorded using a handheld RTK device. In addition, aerial observations were carried out using the DJI Phantom 4 multispectral drone on the morning of the second day to obtain daytime multispectral images, and images of the calibration whiteboard were taken before the flight observation.

[0036] 2) Pix4D Mapper software was used to stitch the multispectral images taken by drones at night and during the day to obtain multispectral orthophotos of two typical areas during the day and at night, with a spatial resolution of 0.25 meters. The daytime multispectral images were calibrated based on a whiteboard to obtain the surface reflectance of the two areas in the R, G, and B bands. Figure 6 The calibration coefficients and formula (1) are used to perform radiometric calibration on the multispectral and panchromatic bands of the SDGSAT-1 satellite remote sensing image, converting the grayscale value into radiometric brightness value.

[0037] 3) The UAV nighttime multispectral images and daytime surface reflectance images of two typical small areas are corrected using Formula 2 to obtain the surface incident radiance.

[0038] 4) Using the measured ground illumination as the dependent variable and the incident radiance of the drone surface at the corresponding position as the independent variable, the surface illumination estimation model was trained using the random forest algorithm. During the training process, the three parameters of the number of decision trees, the depth of the decision trees, and the feature sampling ratio were debugged using the grid search algorithm, and the accuracy was tested through ten-fold cross validation. The optimal model validation scatter plot is shown in Figure 2, the model was applied to the UAV surface incident radiance, and the 0.25-meter resolution surface incident irradiance maps of two typical small areas were obtained (such as Figure 3 as shown).

[0039] 5) The Gram-schmidt orthogonal transform fusion method is used to fuse the three 40m resolution multispectral bands and one 10m resolution panchromatic band of the SDGSAT-1 satellite to obtain three 10m resolution multispectral band images.

[0040] 6) Select the Nanjing Sentinel-2 reflectance product with a similar imaging date to the SDGSAT-1, and extract the surface reflectance of the three visible light bands. Superimpose the 1m resolution land cover classification map of Nanjing on the SDGSAT-1 remote sensing image, and calculate the tree cover ratio and building cover ratio in each 10m resolution pixel. According to the SDGSAT-1 remote sensing image after radiation calibration of Nanjing, the surface reflectance data, building cover ratio and tree cover ratio are corrected by formula (4) to obtain the surface incident radiance remote sensing data of Nanjing.

[0041] 7) Upscale the 0.25-meter surface incident illumination data of two typical small areas obtained by the UAV. During the upscaling process, the average illumination values ​​of non-building and non-vegetation pixels in 1600 0.25-meter resolution pixels in each 10-meter resolution grid are cyclically counted to generate the 10-meter resolution surface incident illumination of two typical small areas. The 10-meter resolution surface incident illumination is used as the dependent variable, and the 10-meter resolution SDGSAT-1 surface radiation brightness remote sensing data obtained after fusion and correction is used as the independent variable. The random forest algorithm is used to train the surface illumination estimation model. During the training process, the three parameters of the number of decision trees, the depth of the decision trees, and the feature sampling ratio are debugged using the grid search algorithm, and the accuracy is tested through ten-fold cross validation. The scatter plot for constructing the optimal model verification is shown in Figure 4 The model was applied to the SDGSAT-1 surface incident radiance, and the nighttime incident surface illumination distribution map of Nanjing was obtained ( Figure 5 ).

[0042] 8) As can be seen from the figure, there are obvious brightness differences in the light environment in Nanjing. The brightness values ​​in urban areas such as the main urban area, Jiangbei New Area and Jiangning District are relatively high, especially in commercial centers and densely populated areas such as Xinjiekou and Confucius Temple. The brightness in the southern part of Jiangning District, the northern part of Liuhe District and ecological protection areas such as Qixia Mountain is relatively low. However, some peripheral residential areas, such as some newly built communities in Jiangning and Liuhe, are not bright enough, which may affect nighttime safety and living convenience. The brightness values ​​in some residential areas in the urban area are relatively high, and there is an obvious light pollution problem.

[0043] 9) The results of this invention show that the urban light environment can be effectively monitored based on ground measured data, UAV data and SDGSAT-1 remote sensing image data, providing technical support for the rational management and control of nighttime urban lighting.

[0044] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0045] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for remote sensing monitoring of urban night light environment using combined unmanned aerial vehicle and SDGSAT-1 satellite data, characterized in that: The method comprises the following steps: Select an observation area, conduct drone aerial photography and field observation, obtain multispectral remote sensing images at night and during the day, and the observed illumination values ​​of the observation points; calibrate the multispectral remote sensing images during the day to obtain the surface reflectance information of the observation area; Combined with the surface reflectivity information of the observation area, the nighttime surface reflected radiance received by the drone is converted into surface incident radiance; The surface incident radiance of the observation point is taken as the independent variable, and the observed illuminance value of the observation point is taken as the dependent variable. The random forest algorithm is used to train the first surface incident illuminance estimation model based on UAV night images. The trained first surface incident illuminance estimation model is applied to the surface incident radiance corresponding to the UAV to obtain the UAV surface incident illuminance of the observation area, and it is upscaled to the same resolution as the SDGSAT-1 satellite remote sensing image. The SDGSAT-1 satellite surface incident radiance of the observation area is calculated based on the SDGSAT-1 satellite remote sensing images after radiation calibration, surface reflectance data, building coverage ratio and tree coverage ratio; The SDGSAT-1 satellite surface incident radiance in the observation area is taken as the independent variable, and the corresponding upscaled UAV surface incident illuminance is taken as the dependent variable. The random forest algorithm is used to train the second surface incident illuminance estimation model based on the SDGSAT-1 satellite. The trained second surface incident illuminance estimation model is applied to the surface incident radiance of the SDGSAT-1 satellite in the entire city, and the night-time incident surface illuminance distribution map of the entire city is obtained.

2. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The process of conducting drone aerial observations and field observations includes the following steps: While the SDGSAT-1 satellite is passing by at night, we select observation areas and use drones equipped with multispectral imagers to conduct flight observations to obtain multispectral remote sensing images at night. During the UAV flight observation, a handheld illuminance meter was used to measure and record the downlink illuminance value of the observation point horizontally upward, and a handheld RTK device was used to record the latitude and longitude position data of the observation point; Use a multispectral imager mounted on a drone to conduct aerial observations in the same observation area during the daytime at a similar time phase, obtain daytime multispectral remote sensing images in the observation area, and take calibration whiteboard images before and after the flight observation; The multispectral remote sensing images at night and during the day are stitched together using UAV orthophoto processing software.

3. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, Combined with the surface reflectivity information of the observation area, the process of converting the surface reflected radiance received by the drone into the surface incident radiance includes the following steps: The data of the three visible light bands of RGB in the multispectral remote sensing images of the drone at night are retained to correct the multispectral remote sensing images of the drone at night; Convert the surface reflected radiance received by the drone into the surface incident radiance: ； In the formula, is the surface incident radiance of the UAV in the i-th band; is the radiation brightness at the entrance pupil of the drone in the i-th band; is the surface reflectivity of the UAV in the i-th band.

4. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The process of using the random forest algorithm to train the first surface incident illumination estimation model based on UAV nighttime images includes: According to the latitude and longitude information of the measured points, the surface incident radiance values ​​of the three visible light bands of RGB in the corresponding multispectral remote sensing images of the UAV at night are extracted; Taking the surface incident radiance of the observation point as the independent variable and the observed illuminance value of the observation point as the dependent variable, the random forest algorithm is used to train the first surface incident illuminance estimation model based on UAV night images. During the training process, grid search is used to optimize the model parameters, and ten-fold cross validation is used to verify the model accuracy.

5. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The process of upscaling the UAV surface incident illumination in the observation area to the same resolution as the SDGSAT-1 satellite remote sensing image includes the following steps: The 10-meter grids are cycled one by one, and the average surface incident illumination of non-building and tree pixels in each 10-meter grid is counted to obtain the surface incident illumination map of the observation area with a 10-meter resolution.

6. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The SDGSAT-1 satellite images were pre-processed by mosaicking and cropping using remote sensing processing software, and radiometric calibration was performed to convert the grayscale values ​​into radiance values: ； In the formula, is the radiance of the satellite at the entrance pupil in the i-th band; is the gray value of the satellite's i-th band; and are the satellite i-th band gain and offset respectively.

7. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The process of calculating the SDGSAT-1 satellite surface incident radiance of the observation area includes the following steps: The three visible light bands of RGB with a resolution of 40m and the panchromatic band with a resolution of 10m in the SDGSAT-1 satellite were fused by Gram-Schmidt orthogonal transformation to obtain the three visible light band images with a resolution of 10m. The surface reflectance data of the three visible light bands of RGB are extracted from the satellite daytime surface reflectance products of the adjacent phase; the building vector data and tree cover data are superimposed on the SDGSAT-1 pixels, and the building coverage and tree cover in each 10-meter resolution pixel are calculated; The SDGSAT-1 satellite surface incident radiance in the observation area is calculated using the following formula: ； In the formula, is the surface incident radiation brightness of the satellite in the i-th band; represents the radiance of the satellite at the entrance pupil in the i-th band; is the surface reflectivity of the satellite’s i-th band; is the building coverage; is the tree cover.

8. The urban night light environment remote sensing monitoring method of the combined unmanned aerial vehicle and SDGSAT-1 satellite data according to claim 1 is characterized in that, The method further comprises: The nighttime surface incident illumination distribution map obtained based on remote sensing represents the urban light environment. The nighttime insufficiently and excessively illuminated areas in the city are identified by combining the surface incident illumination distribution map with the city map and functional zoning.

Images