Urban light pollution identification method by fusing drone images and satellite imagery

Abstract

This invention relates to the field of image processing technology, specifically a method for identifying urban light pollution by fusing UAV images and satellite imagery. The method includes the following steps: S1: fusing UAV images with satellite remote sensing images; S2: performing image analysis and brightness measurement on the obtained fused image; S3: calibrating the measured image brightness information data from the fused image; S4: determining the light pollution risk level based on the image brightness information. This method provides a pathway for large-scale monitoring of the urban light environment. By fusing UAV images with satellite remote sensing images, it enhances image resolution and improves the efficiency of monitoring overall brightness and illuminance data in urban areas. It also improves the accuracy of data acquisition and reduces labor and time costs.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for identifying urban light pollution by fusing drone images and satellite imagery. Background Technology

[0002] Traditional methods for monitoring urban light pollution involve on-site measurements using illuminance meters and luminance meters. Data is then compared to determine and monitor light pollution based on the illuminance and luminance limit levels for different spaces and areas as specified in the "Outdoor Lighting Interference Light Limitation Standard" (GB / T 35626—2017).

[0003] Traditional methods for monitoring light pollution are limited by the capabilities of the monitoring instruments, making it impossible to acquire large-scale regional light environment data. Furthermore, they are susceptible to the influence and interference of different environments and locations, making it impossible for humans to measure every single point within the environment. Data from luminance meters and illuminance meters also exhibit significant errors. They cannot conveniently and macroscopically acquire all light environment data within the visible range of the human eye, thus hindering the analysis of light pollution sources. Therefore, using traditional luminance and illuminance measurement methods requires substantial manpower and time, and the measurement range is limited, resulting in a lack of comprehensiveness in urban light pollution research. Summary of the Invention

[0004] In order to effectively solve the problems in the background art, the present invention proposes a method for urban light pollution identification by fusing UAV images and satellite images.

[0005] The specific technical solution is as follows:

[0006] A method for identifying urban light pollution by fusing drone images and satellite imagery includes the following steps:

[0007] S1: Fusion of UAV imagery and satellite remote sensing effects;

[0008] S2: Perform image analysis and brightness measurement on the obtained fused image;

[0009] S3: Calibrate the image brightness information data measured from the fused image;

[0010] S4: Determine the level of light pollution risk based on image brightness information.

[0011] Preferably, the specific method of S1 is as follows: first, the drone aerial image is denoised; then, the drone aerial image is orthorectified; and finally, the images of the captured areas are stitched together to obtain the drone orthophoto image of the research object. The remote sensing satellite image is preprocessed, including image denoising, orthorectification, geometric fine correction, and image fusion is performed using the HIS image fusion method.

[0012] Preferably, the specific method of S2 is as follows: performing data analysis and measurement on the obtained fused image to obtain the brightness information of the pixels in the analyzed digital image, and generating an analysis diagram from the brightness information of the pixels in the analyzed digital image.

[0013] Preferably, the specific method of S3 is as follows: the ground illuminance is inverted from the image brightness information measured by the fused image, a fitting relationship is established with the ground measured illuminance data using the random forest algorithm, the relationship between the brightness of the fused image and the ground illuminance is calibrated, and the fused image calibrated ground illuminance model can be obtained.

[0014] Preferably, in accordance with the "Specification for Limitation of Interference Light in Outdoor Lighting" (GB / T 35626—2017), the classification of light pollution and the illuminance and brightness are specified, and the ground illuminance model generated by the fused image is used to identify and judge the light pollution risk level of the study area.

[0015] Preferably, the HIS image fusion method converts the low-resolution image RGB model into...

[0016] The HIS model is then used, and the I component is replaced with a high-resolution panchromatic image. Finally, the fused image is obtained through inverse HIS transformation. However, when fusing with aerial imagery, the most appropriate band image must be selected from UAV imagery. The band selection is based on the following formula:

[0017]

[0018] In the formula: I′ represents the band obtained from the UAV imagery; UAV(:,:,K) represents the number of bands in the UAV imagery, where the value of a is determined by calculating the correlation coefficient between the histogram of the multispectral imagery (MSS) and the bands;

[0019] The correlation coefficient is a measure of the degree of linear correlation between variables. It is based on the difference between two variables and their respective means, and is calculated by multiplying the two differences. The formula is as follows:

[0020]

[0021] By calculating the coefficients between each band of UAV imagery and remote sensing imagery, the image bands that need to be fused are determined. α is the cosine angle. The larger α is, the lower the correlation between the two bands. The correlation coefficient index in remote sensing image fusion reflects the spectral preservation between two images.

[0022] Preferably, image parsing needs to be performed in a Python environment. The pixel coordinates, RGB values, grayscale values, and brightness values ​​corresponding to the points are read and recorded as brightness information. The grayscale image is then transformed using a functional relationship to obtain the parsed digital image.

[0023] The functional relationship is: L′(x′, y′)=k′×[α1R′(x′, y′)+α2G(x′, y′)+α3B′(x′, y′)]

[0024] In the formula, L′(x′, y′) is the brightness value of the grayscale image at (x′, y′), R′(x′, y′), G(x′, y′) and B′(x′, y′) are the relationship between the grayscale values ​​of the RGB channels and the actual stimulus values, α1, α2 and α3 are the coefficients corresponding to the RGB channels, and k′ is the second proportionality coefficient.

[0025] Preferably, the light pollution impact assessment result is calculated using the following formula:

[0026]

[0027] In the formula, i represents pedestrians, and j represents light pollution sources in the image. Let L be the light pollution impact of light source j on i, L be the image brightness value corresponding to the light pollution source in the fused image, D be the kernel function bandwidth parameter, and d be the distance between i and j. The cumulative impact of the same type of light source is used as the light pollution assessment object, and the assessment formula is:

[0028] in The threshold for the impact of the same light source on pedestrians is determined by calculating the various light sources in the area and selecting the maximum value. This allows us to identify whether light pollution exists in the area and the main sources of pollution, i.e., the light clusters in the area.

[0029] Compared with existing technologies, the beneficial effects of this invention are: This method provides a way to monitor the urban light environment over a large area. By fusing satellite remote sensing images with UAVs, the image resolution is enhanced, and the efficiency of monitoring the overall brightness and illuminance data of urban areas is improved. This increases the accuracy of data acquisition and reduces labor and time costs.

[0030] By fusing digital image information with ground-based measured data, the correlation between fused image brightness extraction and the ground surface is demonstrated. Using POI data from the Gaode Maps platform and image brightness, combined with a light pollution assessment formula and preset light pollution assessment thresholds, the degree of light pollution in ground areas can be objectively and accurately assessed. Furthermore, by registering satellite map images, urban light pollution can be obtained, providing strong data support for subsequent light pollution control and management. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the present invention. Detailed Implementation

[0032] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways, rotated 90 degrees, or in other orientations, and the spatial relative descriptions used herein will be interpreted accordingly.

[0033] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings and preferred embodiments. Figure 1 As shown, this invention proposes a method for urban light pollution identification by fusing drone images and satellite imagery, comprising the following steps:

[0034] S1: UAV image and satellite remote sensing image fusion technology: This includes an image preprocessing module, which first performs noise reduction and orthorectification on the UAV aerial images, and then stitches the captured areas together to obtain the UAV orthorectified nighttime light image of the research object. Secondly, image preprocessing is performed on the remote sensing satellite images, including image noise reduction, orthorectification, and geometric fine correction. Image fusion is then performed using the HIS image fusion method.

[0035] S2: Fusion Image Data Analysis and Measurement: The obtained fusion image is analyzed and measured to obtain the brightness information of the pixels in the analyzed digital image, and the brightness information of the pixels in the analyzed digital image is used to generate an analysis diagram.

[0036] S3: Image Brightness Information Data Calibration: The ground illuminance is retrieved from the measured image brightness information of the fused image. A fitting relationship is established between the fused image brightness and the measured ground illuminance data using the random forest algorithm, thus calibrating the relationship between the fused image brightness and the ground illuminance. This yields a fused image-calibrated ground illuminance model.

[0037] S4: Identifying Light Pollution Based on Image Brightness Information: According to the "Outdoor Lighting Interference Light Limitation Standard" (GB / T35626—2017), the classification, illuminance, and brightness of light pollution are specified. A ground illuminance model generated from fused images is used to identify and determine the light pollution risk level of the study area.

[0038] To further explain, satellite remote sensing images undergo denoising, geometric calibration, and orthorectification preprocessing; UAV aerial images undergo denoising, orthorectification, and image fusion is performed after image stitching. The fusion process is implemented in MATLAB using the HIS transform method, which converts the low-resolution image RGB model into a more accurate representation of the image.

[0039] The HIS model is then used, and the I component is replaced with a high-resolution panchromatic image. Finally, an inverse HIS transform is performed to obtain the fused image. However, when fusing with aerial imagery, the most appropriate band image must be selected from UAV imagery. The band selection is based on the following formula:

[0040]

[0041] In the formula: I′ represents the band acquired from the UAV imagery; UAV(:,:,K) represents the number of bands in the UAV imagery. The value of a is determined by calculating the correlation coefficient between the histogram of the multispectral imagery (MSS) and the bands.

[0042] The correlation coefficient is a measure of the degree of linear correlation between variables. It is based on the difference between two variables and their respective means, and is calculated by multiplying these two differences. The formula used in this paper is as follows:

[0043] By calculating the coefficients between different bands of UAV imagery and remote sensing imagery, the required fusion is determined.

[0044] Image bands. α is the cosine angle; the larger α is, the lower the correlation between the two bands. In remote sensing image fusion, the correlation coefficient reflects the spectral preservation between two images.

[0045] Based on the HIS fusion principle, only one band is needed to replace I. The selected UAV image band replaces the original I image to obtain the I′ image. Finally, an inverse HIS transform is performed on H, S, and I′ to obtain the fused image.

[0046] Furthermore, image analysis and brightness measurement are performed on the obtained fused image. Image analysis needs to be performed in a Python environment, reading and recording the pixel coordinates, RGB values, grayscale values, and brightness values ​​corresponding to the points. The grayscale image is then transformed using a functional relationship to obtain the analyzed digital image.

[0047] The functional relationship is

[0048] L′(x′,y′)=k′×[α1R′(x′,y′)+α2G(x′,y′)+α3B′(x′,y′)]

[0049] In the formula, L′(x′, y′) is the brightness value of the grayscale image at (x′, y′), R′(x′, y′), G′(x′, y′) and B′(x′, y′) are the relationship between the grayscale values ​​of the RGB channels and the actual stimulus values, α1, α2 and α3 are the coefficients corresponding to the RGB channels, and k′ is the second proportionality coefficient.

[0050] The derivation of the functional relationship is as follows:

[0051] Establish a functional relationship between the grayscale values ​​of the three RGB channels of a pixel and the stimulus values ​​in the RGB color space;

[0052] By establishing the functional relationship between the grayscale values ​​of the three RGB channels and the stimulus values ​​of the RGB color space, as well as the conversion relationship between the 1931CIE-RGB system and the 1931CIE-XYZ system, the relationship between the grayscale values ​​of the three RGB channels and the actual brightness values ​​can be successfully established, thus obtaining the final functional relationship.

[0053] The grayscale value DN of the merged image file (.tif) can convert the image grayscale to spectral radiance:

[0054] L i =DN·gain i +offset i

[0055] Among them, L i Let DN be the zenith spectral radiance corresponding to the gray value DN of image band i, and gain i and offset i The absolute radiometric correction coefficients for spectral band i are stored in the satellite image metadata file (_meta.xml). i For gain, offset i For bias.

[0056] The formula for retrieving illuminance is: L = R × E

[0057] In the formula, L represents luminance; R represents the reflectance coefficient; and E represents illuminance. Thus, ground illuminance is obtained through luminance.

[0058] The study area in the analyzed image is measured and data is read using a uniformly distributed grid. A measurement grid is equidistantly arranged on the analyzed digital image. The margins of the measurement grid in the four directions (top, bottom, left, and right), as well as the required number of rows and columns of points, are customized. The system intelligently generates an equidistantly distributed measurement grid and outputs the parameters and average values ​​for each point.

[0059] The principle of the measurement grid layout is as follows:

[0060] L1 = Lp -(L left -L right )

[0061] H1 = H p -(H up -H down )

[0062] In the formula, L1 is the length of the entire grid, H1 is the height of the entire grid, and L left L is the left margin. right H is the right margin. up H is the top margin. down This is the bottom margin. The overall grid layout can be determined based on the input in the prompt box, in pixels.

[0063] Furthermore, the image brightness data was established using the random forest algorithm, and the relationship between the image brightness inversion ground illuminance data and the measured ground brightness illuminance data was verified to validate the degree of correlation.

[0064] Furthermore, the study area is divided, and the RGB values ​​of the digital images are substituted into the photoelectric conversion function to collect the generated values. Appropriate charts are then selected as needed. The average brightness and illuminance values ​​of the images in the study area are calculated, yielding the overall basic light environment data for the region.

[0065] Furthermore, the number of points of interest (POIs) in the study area was obtained through the Gaode Map platform, where pedestrians within the area were considered POIs, uniformly representing the activity density of the area. It is generally believed that areas with a high number of POIs are more adaptable to ambient brightness, while areas with a low number of POIs do not require excessive lighting. To estimate whether light pollution impact exists in the study area, a kernel density estimation method (KDE) is provided, where each point has its own kernel function to estimate its impact on the surrounding area. Furthermore, the impact rate decreases smoothly with distance, determined by the bandwidth parameter. There is also an attribute parameter; as this parameter increases, the impact rate increases at the point's location and further expands with distance. When estimating the impact of light pollution sources on their surrounding areas, we use the kernel function as the impact simulation function, and the L value as the attribute value of the function. The light pollution impact assessment results can be calculated by the following formula:

[0066] In the formula, i represents pedestrians, and j represents light pollution sources in the image. Let L be the light pollution impact of light source j on i. Let L be the image brightness value corresponding to the light pollution source in the fused image, D be the kernel function bandwidth parameter, and d be the distance between i and j. This paper uses the cumulative impact of the same type of light source as the light pollution assessment object, and the assessment formula is:

[0067] in The threshold for the impact of the same light source on pedestrians is determined by calculating the various light sources in the area and selecting the maximum value. This allows us to identify whether light pollution exists in the area and the main sources of pollution, i.e., the light clusters in the area.

[0068] The specific implementation steps are as follows.

[0069] Step 1: Image Acquisition and Preprocessing

[0070] Satellite nighttime light images were captured on a specific day; these images can be found on the Jilin-1 website. Aerial drone images of the study area were also taken at night, ensuring clear, cloudless weather and good air quality on the day the images were acquired. The satellite remote sensing images underwent noise reduction, orthorectification, and geometric calibration preprocessing. The drone images also underwent noise reduction, orthorectification, and image stitching preprocessing. This resulted in preprocessed remote sensing and orthorectified aerial images.

[0071] Step 2: Image Fusion

[0072] The HIS image transformation method is used to fuse satellite remote sensing images and UAV aerial images. Image fusion can be implemented in the MATLAB environment. During image fusion, the selected UAV image bands replace the original I image to obtain the I' image. Finally, an inverse HIS transformation is performed on H, S, and I' to obtain the fused image. The most appropriate UAV image bands are selected based on the following formula:

[0073]

[0074] In the formula: I' represents the band acquired from the UAV imagery; UAV(:,:,K) represents the number of bands in the UAV imagery. The value of a is determined by calculating the correlation coefficient between the histogram of the multispectral imagery (MSS) and the bands.

[0075] Step 3: Image Analysis

[0076] The fused satellite image undergoes brightness analysis. The analysis environment, implemented in Python, reads and records the pixel coordinates, RGB values, grayscale values, and brightness values ​​corresponding to each point as brightness information. The grayscale image is converted using a functional relationship to obtain the analyzed digital image. The grayscale value DN of the fused image file (.tif) can be converted from image grayscale to spectral radiance.

[0077] L i =DN·gain i +offset i

[0078] Among them, L iLet DN be the zenith spectral radiance corresponding to the gray value DN of image band i, and gain i and offset i The absolute radiometric correction coefficients for spectral band i are stored in the satellite image metadata file (_meta.xml). i For gain, offset i For bias.

[0079] Step 4: Image Calibration

[0080] To calibrate the analyzed image data, the study area first needs to be gridded to generate an equidistant grid. The average brightness and illuminance at each point are then calculated. A random forest model is used to derive a correlation model between the calculated image brightness and the measured brightness of the feature regions. Finally, image calibration is performed on the ground brightness and illuminance.

[0081] Step 5: Identification and Judgment

[0082] By obtaining pedestrian points of interest (POIs) in the identified area using Amap (Gaode Maps), and assessing the area's light pollution risk using the kernel density estimation method (KDE), the evaluation result can be calculated using the following formula:

[0083]

[0084] In the formula, i represents pedestrians, and j represents light pollution sources in the image. Let L be the light pollution effect of j on i. Let L be the image brightness value corresponding to the light pollution source in the fused image, D be the kernel function bandwidth parameter, and d be the distance between i and j. Using the light pollution evaluation formula and a preset light pollution evaluation threshold, the threshold evaluation formula is:

[0085]

[0086] in The threshold for the impact of the same light source on pedestrians is determined by image data discrimination calculation, which can objectively and accurately determine whether light pollution exists and assess the degree of light pollution.

[0087] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for identifying urban light pollution by fusing drone images and satellite imagery, characterized in that: Includes the following steps: S1: Fusion of UAV images and satellite remote sensing images; S2: Perform image analysis and brightness measurement on the obtained fused image; S3: Calibrate the image brightness information data measured from the fused image; S4: Determine and identify the level of light pollution risk based on image brightness information; The specific method of S1 is as follows: first, the drone aerial image is denoised, then the drone aerial image is orthorectified, and the captured area is stitched together to obtain the drone orthorectified night light image of the research object; the remote sensing satellite image is preprocessed, including image denoising, orthorectification, geometric fine correction, and image fusion is performed using the HIS image fusion method. The HIS image fusion method converts the low-resolution image RGB model into an HIS model, then replaces the I component with a high-resolution panchromatic image, and finally performs an inverse HIS transformation to obtain the fused image. However, when fusing with aerial imagery, the most appropriate band image must be selected from the UAV imagery. The band selection is based on the following formula: ; In the formula: I′ represents the band obtained from the UAV imagery; UAV(:,:,K) represents the number of bands in the UAV imagery, where the value of a is determined by calculating the correlation coefficient between the histogram of the multispectral image MSS and the bands; The correlation coefficient is a measure of the degree of linear correlation between variables. It is based on the difference between two variables and their respective means, and is calculated by multiplying the two differences. The formula is as follows: ; By calculating the coefficients between each band of UAV imagery and remote sensing imagery, the image bands that need to be fused are determined. α is the cosine angle. The larger α is, the lower the correlation between the two bands. The correlation coefficient index in remote sensing image fusion reflects the spectral preservation between two images.

2. The urban light pollution identification method based on the fusion of UAV images and satellite imagery according to claim 1, characterized in that, The specific method of S2 is as follows: data analysis and measurement are performed on the obtained fused image to obtain the brightness information of the pixels in the analyzed digital image, and the brightness information of the pixels in the analyzed digital image is used to generate an analysis diagram.

3. The urban light pollution identification method based on the fusion of UAV images and satellite imagery according to claim 1, characterized in that, The specific method of S3 is as follows: the ground illuminance is inverted from the image brightness information measured by the fused image, and a fitting relationship is established with the ground measured illuminance data using the random forest algorithm. The relationship between the brightness of the fused image and the ground illuminance is calibrated, and the fused image calibrated ground illuminance model can be obtained.

4. The urban light pollution identification method based on the fusion of UAV images and satellite imagery according to claim 1, characterized in that, According to the "Specification for Limitation of Interference Light in Outdoor Lighting" (GB / T 35626—2017), the classification of light pollution and the illuminance and brightness are specified. The ground illuminance model generated by the fused image is used to identify and judge the light pollution risk level of the study area.

5. The urban light pollution identification method based on the fusion of UAV images and satellite imagery according to claim 1, characterized in that, Image parsing needs to be done in a Python environment. The pixel coordinates, RGB values, grayscale values, and brightness values ​​corresponding to the points are read and recorded as brightness information. The grayscale image is then transformed using a function to obtain the parsed digital image. The functional relationship is: ; In the formula, For grayscale images in The brightness value at that location, , and These are the formulas relating the grayscale values ​​of the RGB channels to the actual stimulus values. These are the coefficients corresponding to the RGB channels, and k' is the second proportional coefficient.

6. The urban light pollution identification method based on the fusion of UAV images and satellite imagery according to claim 3, characterized in that, The light pollution impact assessment result is calculated using the following formula: ; In the formula, i represents pedestrians, and j represents light pollution sources in the image. Let L be the light pollution impact of j on i, L be the image brightness value corresponding to the light pollution source in the fused image, D be the kernel function bandwidth parameter, and d be the distance between i and j. The cumulative impact of the same type of light source is used as the light pollution assessment object, and the assessment formula is: ; in The threshold for the impact of the same light source on pedestrians is determined by calculating the various light sources in the area and selecting the maximum value. This allows us to identify whether light pollution exists in the area and the main sources of pollution, i.e., the light clusters in the area.

Images