A remote sensing image feature recognition method
By screening remote sensing images and fusion of drone data, combined with visual and geomorphological similarity analysis, the difference problem in remote sensing image feature recognition is solved, simplifying the acquisition process and improving the efficiency of geological survey.
Patent Information
- Application Number
- CN202410907870.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-08
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-08
AI Technical Summary
In the prior art, remote sensing image feature recognition has differential problems caused by differences in satellite data provision platforms, and the drone acquisition steps are cumbersome and time-consuming.
By registering and logging into the satellite data platform to download geological remote sensing images, filtering and registering images, using drones to obtain image data, perform visual and geomorphological similarity analysis, calculate evaluation index, select optimal images, and perform image fusion based on evaluation values and Supplementary collection and optimize route planning to improve collection efficiency.
It realizes more accurate remote sensing image feature recognition, simplifies the drone acquisition process, and improves the efficiency of geological surveys and the integrity of images.
Smart Images

Figure CN118537759B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of remote sensing image feature recognition, and in particular to a remote sensing image feature recognition method. Background Art
[0002] Remote sensing image feature recognition method is a technology that uses remote sensing images and technology to analyze and identify geological features. This method can effectively assist geologists and mineral explorers in identifying important geological information in different geological environments.
[0003] However, due to the technical specifications of satellites, sensor types, data acquisition time and conditions, and data processing and generation methods, different satellite data providing platforms have led to differences in the remote sensing images collected for the geology of the target area. In the existing technology, the remote sensing images are directly obtained through the satellite data providing platform, or large-scale drone collection work is carried out in the geological survey area. The former may lead to differences in remote sensing image features, while the latter has more complicated steps to obtain geological survey features, which is time-consuming and laborious.
[0004] Therefore, a remote sensing image feature recognition method is proposed. Summary of the invention
[0005] In view of this, the present invention provides a remote sensing image feature recognition method to solve the problems raised by the above background technology.
[0006] The purpose of the present invention can be achieved through the following technical solutions:
[0007] Obtain remote sensing images: Obtain remote sensing images: Register and log in to the corresponding account to access different satellite data provision platforms, download geological remote sensing images of the target area; and conduct preliminary screening of the downloaded geological remote sensing images based on the comparison results of the numerical evaluation values of each satellite data provision platform;
[0008] Aerial photography data acquisition: register the selected preset images, identify and extract the overlap range of the preset images, and randomly select a set range from the obtained overlap range as a reference local image; extract the reference local image corresponding to the selected range of the current preset images; then use the drone to take aerial photos of the selected set range to obtain the drone image data of the selected set range;
[0009] Image similarity analysis: convert the UAV image data and the corresponding reference local image into GeoTIFF format; preprocess the converted GeoTIFF images; after preprocessing, analyze the visual similarity between the UAV image data and the corresponding reference local image to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image;
[0010] Geomorphic similarity analysis: extract the digital elevation model DEM from the reference local image corresponding to the current target area, recorded as the reference local image DEM, and extract the digital elevation model DEM from the drone image data, recorded as the drone DEM; pre-process the extracted DEM data; analyze the geomorphic similarity between the drone image data and the corresponding reference local image, and obtain the geomorphic similarity evaluation index SUG of the current target area corresponding to the reference local image;
[0011] Select remote sensing images: Based on the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG of the reference local image corresponding to the current target area; according to the formula Perform normalized calculation to obtain the selected evaluation value XQP of the reference local image corresponding to the current target area; where gv1 and gv2 are the influencing weight factors of the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG respectively;
[0012] Arrange the selected evaluation values XQP corresponding to each reference local image of the current target area in descending order, and select the reference local image with the largest evaluation value XQP as the marked reference local image; extract the preset image corresponding to the marked reference local image, and use the preset image as the final remote sensing image for feature recognition of the current target area;
[0013] Local area improvement: Based on the overlapping range between the preset images, the final remote sensing image of the current target area is segmented into the overlapping range area, and the remaining area except the overlapping range area is used as the remaining supplementary area of the final remote sensing image. The corresponding route with the largest collection efficiency is selected by the drone to collect and supplement the remaining supplementary area of the current target area, and the remaining supplementary area image collected by the drone is fused with the final remote sensing image to generate a complete remote sensing image of the current target area.
[0014] In some embodiments, the visual similarity between the drone image data and the corresponding reference local image is analyzed, and the specific analysis steps are:
[0015] Extract reflectance values from each band of the drone image data of the current target area and the corresponding reference local image; calculate the reflectance difference for each pixel point of the drone image data and the corresponding reference local image, that is, calculate by the formula reflectance difference value (i, j) = WX (i, j) - WR (i, j), and take the absolute value of the calculated result; where i and j represent the row and column coordinates of the pixel point, respectively, WX represents the reflectance value of the corresponding reference local image, and WR represents the reflectance value of the drone image data;
[0016] Extract each group of reflection difference values of the reference local image corresponding to the current target area, calculate the mean, and use the calculated mean as the difference mean of the reference local image corresponding to the current target area; calculate each group of reflection difference values of the reference local image corresponding to the current target area using the standard deviation formula, and use the calculated result as the difference dispersion value of the reference local image corresponding to the current target area; extract the maximum reflection difference value and the minimum reflection difference value from each group of reflection difference values of the reference local image corresponding to the current target area, calculate the difference, and use the calculated difference as the maximum difference of the reference local image corresponding to the current target area;
[0017] The difference mean, difference dispersion and difference maximum values corresponding to the reference local image of the current target area are multiplied by the corresponding preset weight coefficients and then summed to obtain the difference evaluation index CY1 of the reference local image of the current target area.
[0018] In some embodiments, the specific steps of obtaining the visual similarity evaluation index SJP of the reference local image corresponding to the current target area are:
[0019] The reference local image and drone image data corresponding to the current target area are processed through the GLCM tool in the image processing software, and the same GLCM parameters are set; the frequency distribution of each pixel pair of the reference local image and drone image data corresponding to the current target area is calculated respectively to generate a GLCM matrix; the contrast, homogeneity, energy and entropy parameters are extracted from each generated GLCM matrix; the ratio of the reference local image corresponding to the current target area and the drone image data in terms of contrast, homogeneity, energy and entropy parameters is calculated respectively, and the difference between the ratio calculated for each parameter and the integer 1 is calculated, and the absolute value is taken; the mean of each group of differences after taking the absolute value of the reference local image corresponding to the current target area is calculated, and the mean is used as the texture evaluation index CY2 of the reference local image corresponding to the current target area;
[0020] Identify and extract the edge information of the reference local image and drone image data corresponding to the current target area, and extract the contour from the edge information; for the two extracted contours, calculate the area corresponding to the two contours; calculate the Hu moment for the two extracted contours, and divide the Hu moment eigenvalues corresponding to the two contours by the corresponding contour area to obtain the 7 normalized Hu moments corresponding to the reference local image and drone image data of the current target area; calculate the ratio of the 7 normalized Hu moments of the reference local image corresponding to the current target area to the 7 normalized Hu moments of the drone image data one by one; calculate the difference between the ratio calculated by each parameter and the integer 1, and take the absolute value; calculate the mean of each group of differences after taking the absolute value of the reference local image corresponding to the current target area, and use the mean as the contour evaluation index CY3 of the reference local image corresponding to the current target area;
[0021] Substitute the difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3 obtained by analyzing the reference local image corresponding to the current target area into the formula Normalized calculation is performed to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image; where CY1 参考 CY2 参考 and CY3 参考 are the set reference difference evaluation index, reference texture evaluation index and reference contour evaluation index respectively; zy1, zy2 and zy3 are the influencing weight factors of the difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3 respectively, and their values are set to 1.235, 1.217 and 1.228 respectively.
[0022] In some embodiments, the topographic similarity between the drone image data and the corresponding reference local image is analyzed, and the specific analysis steps are:
[0023] Extract corresponding terrain characteristic values from the drone image DEM and the corresponding reference local image DEM; the terrain characteristic values include slope, aspect, curvature and altitude;
[0024] The slope characteristic value of each pixel point of the reference local image DEM corresponding to the current target area and the slope characteristic value of the pixel point of the drone image DEM are calculated to obtain the slope similarity value PM of the reference local image DEM and the drone image DEM corresponding to the current target area; that is, through the formula Calculated; where Sft and Sut represent the slope characteristic values of the drone image DEM and the corresponding reference local image DEM at pixel t, respectively; where t = 1, 2 or E, and E is the total number of pixels;
[0025] The same steps as above are used to calculate the aspect similarity value, curvature similarity value and altitude similarity value of the reference local image DEM and the drone image DEM at each pixel point in the current target area, and they are marked as PN, PC and PZ respectively.
[0026] In some embodiments, the specific steps of obtaining the reference local image topography similarity assessment index SUG corresponding to the current target area are:
[0027] According to the formula Normalization processing is performed to obtain the topographic similarity assessment index SUG of the current target area corresponding to the reference local image; PM 参考 PN 参考 、PC 参考 and PZ 参考are the set reference slope similarity value, reference aspect similarity value, reference curvature similarity value and reference altitude similarity value respectively; ce1, ce2, ce3 and ce4 are the influence weight factors of slope similarity value PM, aspect similarity value PN, curvature similarity value PC and altitude similarity value PZ respectively, and their values are set to 1.085, 1.081, 1.084 and 1.087 respectively.
[0028] In some embodiments, the specific steps of obtaining the collection efficiency value CJX of each route of the drone are:
[0029] T1: Select a drone that meets the current collection mission requirements;
[0030] T2: Plan the UAV route based on the geographical location, area size and terrain characteristics of the remaining supplementary area, and obtain multiple planned routes for the current UAV to the remaining supplementary area;
[0031] T3: Based on the planned multiple routes, calculate the total flight distance required for each route;
[0032] T4: Match the obtained total distance of each drone with the set distance reference range corresponding to the estimated flight time and the power consumption required for the flight; obtain the estimated flight time and the power consumption required for the flight of the drones corresponding to the remaining supplementary area;
[0033] T5: Collect wind speed and wind direction data of the current target area, calculate the angle between each flight route of the drone and the wind direction of the current target area, obtain the wind direction angle value of each route of the drone, and integrate the wind speed and corresponding wind direction angle value of each route into a meteorological influence set, and match each meteorological influence set obtained with each preset meteorological influence set; set each preset meteorological influence set to correspond to a meteorological additional value; obtain the meteorological additional value of the drone on each route;
[0034] T6: Match the meteorological added value of each current route with the meteorological added value reference range corresponding to the set required additional time and required power consumption; obtain the required additional time and required power consumption of each current route, add the estimated flight time and required flight power consumption of each route in step T4 to the corresponding additional time and additional power consumption, and obtain the total flight time and total flight power of the drone on each route;
[0035] T7: Based on the comparison results of the total flight power of each route and the corresponding total power capacity of the drone, the required charging times of the drone on each route are obtained;
[0036] The total distance of the target route, the total flight time, and the number of required charges are multiplied by the corresponding preset weight coefficients and then summed to obtain result one. The collection efficiency value of the target route drone is obtained by dividing the integer 1 by result one.
[0037] In some embodiments, according to the comparison results of the numerical superiority evaluation values SY of each satellite data providing platform, the downloaded remote sensing images are preliminarily screened, and the specific screening steps are:
[0038] The spatial resolution, spectral resolution and temporal resolution of each satellite data provider platform are obtained; the spatial resolution, spectral resolution and temporal resolution of the corresponding satellite data provider platform are multiplied by the corresponding preset weight coefficients and then summed to obtain the numerical quality evaluation value of the corresponding satellite data provider platform; a threshold value of the numerical quality evaluation value is set, and the numerical quality evaluation value of each satellite data provider platform is compared with the set threshold value. If the numerical quality evaluation value of the corresponding satellite data provider platform is greater than the set threshold value, the geological remote sensing image downloaded by the platform is used as the preset image.
[0039] Compared with the prior art, the present invention has the following beneficial effects:
[0040] The present invention obtains the visual similarity evaluation index and the geomorphic similarity evaluation index of the reference local image corresponding to the current target area by analyzing the similarity from the two directions of vision and geomorphology of the image, and further normalizes the visual similarity evaluation index and the geomorphic similarity evaluation index of each reference local image to obtain the selection evaluation value of each reference local image, which can more comprehensively reflect the similarity between each reference local image and the drone image data, provide a basis for the subsequent selection of specific remote sensing images, and improve the accuracy of geological feature recognition;
[0041] The present invention obtains the remaining supplementary area and scientifically plans the UAV route based on the remaining supplementary area, thereby solving the relatively cumbersome problem of the prior art that large-scale UAV collection work is required in the geological survey area, and realizes the targeted collection and supplement of the remaining supplementary area based on the selected image; at the same time, a comprehensive analysis is performed on the total distance, total flight time and number of required charging times of each planned route, which can comprehensively evaluate the collection efficiency of the UAV on the corresponding route, thereby improving the collection efficiency of the UAV for the remaining supplementary area. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:
[0043] Figure 1 It is a principle block diagram of the present invention. DETAILED DESCRIPTION
[0044] Several embodiments of the present application will be described in more detail below with reference to the accompanying drawings so that those skilled in the art can implement the present application. The present application can be embodied in many different forms and purposes and should not be limited to the embodiments described herein. These embodiments are provided to make the present application comprehensive and complete, and to fully convey the scope of the present application to those skilled in the art. The embodiments do not limit the present application.
[0045] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and / or this specification, and will not be interpreted in an idealized or overly formal sense, unless explicitly defined as such herein.
[0046] See also Figure 1 As shown, a remote sensing image feature recognition method includes:
[0047] Obtain remote sensing images: Register and log in to the corresponding account to access different satellite data provision platforms; such as USGSEarthExpl orer, NASA LAADSDAAC, etc.; download geological remote sensing images of the target area;
[0048] According to the comparison results of the numerical quality evaluation values of each satellite data provider platform, the downloaded remote sensing images are preliminarily screened;
[0049] The specific steps to obtain the numerical evaluation value SY of each satellite data providing platform are as follows:
[0050] The spatial resolution, spectral resolution and temporal resolution of each satellite data provider platform are obtained; the spatial resolution, spectral resolution and temporal resolution of the corresponding satellite data provider platform are multiplied by the corresponding preset weight coefficients and then summed to obtain the numerical quality evaluation value of the corresponding satellite data provider platform; a threshold value of the numerical quality evaluation value is set, and the numerical quality evaluation value of each satellite data provider platform is compared with the set threshold value. If the numerical quality evaluation value of the corresponding satellite data provider platform is greater than the set threshold value, the geological remote sensing image downloaded by the platform is used as the preset image;
[0051] Perform preliminary screening based on the comparison results of each platform to obtain the preset images corresponding to the current target area;
[0052] It should be noted that if the numerical excellence evaluation values of each satellite data provider platform are all less than the set threshold, the set threshold will be initially lowered based on the preset downward adjustment range before comparison; similarly, the preset downward adjustment range will be adjusted again based on the set threshold after the initial downward adjustment until a numerical excellence evaluation value greater than the set threshold appears, so that intelligent adjustments can be made according to actual conditions.
[0053] Aerial photography data acquisition: register the selected preset images, use the overlap analysis tool in the image processing software to identify and extract the overlap range of the preset images, and randomly select a set range from the obtained overlap range as a reference local image; extract the reference local image corresponding to the selected range of the current preset images; then use the drone to take aerial photos of the selected set range to obtain the drone image data of the selected set range;
[0054] Image similarity analysis: Use image processing software to convert drone image data and corresponding reference local images into GeoTIFF format; image processing software such as GDAL, QGIS and ArcGIS; pre-process the converted GeoTIFF images; pre-processing includes but is not limited to radiation correction, atmospheric correction and geometric correction;
[0055] After preprocessing, the visual similarity between the UAV image data and the corresponding reference local image is analyzed and calculated to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image;
[0056] The specific calculation steps for the visual similarity evaluation index SJP are:
[0057] K1: Using image processing software, extract reflectance values from each band of the drone image data and the corresponding reference local image;
[0058] For each pixel point of the drone image data and the corresponding reference local image, the reflectivity difference is calculated, that is, the reflection difference value (i, j) = WX (i, j) - WR (i, j) is calculated, and the absolute value of the calculation result is taken; where i and j represent the row and column coordinates of the pixel point, respectively, WX represents the reflectivity value of the corresponding reference local image, and WR represents the reflectivity value of the drone image data;
[0059] Extracting each group of reflection difference values of the current target area corresponding to the reference local image, and calculating the mean, and taking the calculated mean as the difference mean of the current target area corresponding to the reference local image;
[0060] The standard deviation formula is used to calculate the reflection difference values of each group of the current target area corresponding to the reference local image, and the calculation result is used as the difference dispersion value of the current target area corresponding to the reference local image;
[0061] Extract the maximum reflection difference value and the minimum reflection difference value from each group of reflection difference values of the current target area corresponding to the reference local image, calculate the difference, and use the calculated difference as the maximum difference of the reference local image corresponding to the current target area;
[0062] The difference mean, difference dispersion and difference maximum values corresponding to the reference local image of the current target area are multiplied by the corresponding preset weight coefficients and then summed to obtain the difference evaluation index CY1 of the reference local image of the current target area;
[0063] It should be noted that the calculation of the reflection difference value between each reference local image and the drone image data can be used to evaluate the similarity between each reference local image and the drone image data. The larger the reflection difference value, the lower the similarity between the two images. Further analysis of the mean, dispersion and maximum difference based on each pixel point can be used to evaluate the overall similarity between the two images, thereby improving the accuracy of data analysis.
[0064] K2: Use the GLCM tool in the image processing software to set the same GLCM parameters for the reference local image and drone image data corresponding to the current target area, including distance, angle, and grayscale; and calculate the frequency distribution of each pixel pair corresponding to each reference local image and drone image data of the current target area to generate a GLCM matrix;
[0065] Extract statistical parameters from each generated GLCM matrix; the statistical parameters include contrast, homogeneity, energy and entropy parameters;
[0066] Calculate the ratio of the reference local image corresponding to the current target area and the drone image data in terms of contrast, homogeneity, energy and entropy parameters respectively, and perform difference calculation between the ratio calculated by each parameter and the integer 1, and take the absolute value;
[0067] The average of each group of differences after taking the absolute value of the reference local image corresponding to the current target area is calculated, and the average is used as the texture evaluation index CY2 of the reference local image corresponding to the current target area;
[0068] It should be noted that the smaller the calculated mean, the more similar the texture features between the reference local image and the drone image data are.
[0069] K3: Use edge detection algorithm to identify and extract edge information of the current target area corresponding to the reference local image and drone image data, and extract the contour from the edge information;
[0070] For the two extracted contours, use the function in the image processing library to calculate the areas corresponding to the two contours;
[0071] The Hu moments of the two extracted contours are calculated using the functions in the image processing library, and the Hu moment eigenvalues corresponding to the two contours are divided by the corresponding contour areas to obtain the 7 normalized Hu moments corresponding to the reference local image and the drone image data of the current target area; the 7 Hu moments include 3 positive moments and 4 anti-moments;
[0072] Calculate the ratio of the 7 normalized Hu moments of the reference local image corresponding to the current target area to the 7 normalized Hu moments of the drone image data; and calculate the difference between the ratio calculated by each parameter and the integer 1, and take the absolute value;
[0073] The average of each group of differences after taking the absolute value of the current target area corresponding to the reference local image is calculated, and the average is used as the contour evaluation index CY3 of the corresponding reference local image;
[0074] It should be noted that the ratios of all seven Hu moment features are comprehensively evaluated, and the calculated mean reflects the similarity between the contours of each reference local image and the UAV image data. The smaller the mean, the higher the contour similarity.
[0075] K4: Substitute the difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3 obtained by analyzing the reference local image corresponding to the current target area into the formula Normalized calculation is performed to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image; where CY1 参考 CY2 参考 and CY3 参考 are the reference difference evaluation index, reference texture evaluation index and reference contour evaluation index respectively; zy1, zy2 and zy3 are the influence weight factors of difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3 respectively, and their values are set to 1.235, 1.217 and 1.228 respectively;
[0076] It should be noted that the reference local images and UAV image data are analyzed and calculated from three directions: spectrum, texture and contour features, to obtain the difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3. Based on the above three indexes, the similarity between each reference local image and the UAV image data can be reflected from different directions. The above three indexes are further normalized to obtain the visual similarity evaluation index SJP of each reference local image. The normalized visual similarity evaluation index SJP provides a unified evaluation standard, which makes it possible to intuitively compare the similarity between each reference local image and the UAV image data, which is convenient for decision-making and further analysis.
[0077] Geomorphic similarity analysis: Use remote sensing data processing software to extract digital elevation model DEM from the corresponding reference local image, recorded as reference local image DEM, and use stereo image pair analysis method to extract digital elevation model DEM from drone image data, recorded as drone DEM; preprocess the extracted DEM data; preprocessing includes filtering and smoothing to reduce the impact of noise and outliers; preprocessing also includes checking spatial resolution and projection coordinate system, and resampling or coordinate system conversion;
[0078] The geomorphic similarity between the UAV image data and the corresponding reference local image is analyzed and calculated to obtain the geomorphic similarity evaluation index SUG of the current target area corresponding to the reference local image;
[0079] The specific calculation steps for the landform similarity assessment index SUG are as follows:
[0080] V1: Extract corresponding terrain feature values from UAV image DEM and corresponding reference local image DEM; terrain feature values include slope, aspect, curvature and altitude;
[0081] V2: Calculate the slope characteristic value of each pixel point of the reference local image DEM corresponding to the current target area and the slope characteristic value of the pixel point of the drone image DEM to obtain the slope similarity value PM of the reference local image DEM and the drone image DEM corresponding to the current target area;
[0082] That is, through the formula Calculated; where Sft and Sut represent the slope characteristic values of the drone image DEM and the corresponding reference local image DEM at pixel t, respectively; where t = 1, 2 or E, and E is the total number of pixels;
[0083] V3: Similarly, step V2 calculates the slope similarity value, curvature similarity value and altitude similarity value of the reference local image DEM and the drone image DEM at each pixel of the current target area, and marks them as PN, PC and PZ respectively;
[0084] According to the formula Normalization processing is performed to obtain the topographic similarity assessment index SUG of the current target area corresponding to the reference local image; PM 参考 PN 参考 、PC 参考 and PZ 参考are the set reference slope similarity value, reference aspect similarity value, reference curvature similarity value and reference altitude similarity value respectively; ce1, ce2, ce3 and ce4 are the influence weight factors of slope similarity value PM, aspect similarity value PN, curvature similarity value PC and altitude similarity value PZ respectively, and their values are set to 1.085, 1.081, 1.084 and 1.087 respectively;
[0085] It should be noted that the aspect is an angle value, usually in the range of -180-180 degrees; when calculating, it should be ensured that the aspect values of the two DEMs are in the same range, which can be achieved by converting the difference between them into the range of 0-360 degrees for direct comparison.
[0086] It should be noted that the geomorphic similarity between the UAV image DEM and each reference local image DEM is analyzed from multiple terrain features such as slope, aspect, curvature and altitude, and normalized to obtain the geomorphic similarity assessment index SUG of each reference local image. The smaller the geomorphic similarity assessment index SUG, the higher the geomorphic similarity between the reference local image and the UAV image data, thereby providing an accurate similarity measurement and more intuitively comparing the similarity between each reference local image and the UAV image data from the geomorphic direction.
[0087] Select remote sensing images: Based on the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG of the reference local image corresponding to the current target area, normalization processing is performed to obtain the selection evaluation value XQP of the reference local image corresponding to the current target area; specifically:
[0088] According to the formula The normalized calculation is performed to obtain the selected evaluation value XQP of the corresponding reference local image; where gv1 and gv2 are the influence weight factors of the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG, and their values are set to 1.538 and 1.569 respectively;
[0089] Arrange the selected evaluation values XQP corresponding to each reference local image of the current target area in descending order, and select the reference local image with the largest evaluation value XQP as the marked reference local image; extract the preset image corresponding to the marked reference local image, and use the preset image as the final remote sensing image for feature recognition of the current target area;
[0090] It should be noted that by normalizing the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG of each reference local image, the selection evaluation value XQP of each reference local image is obtained, which can more comprehensively reflect the similarity between each reference local image and the UAV image data, thereby improving the accuracy and basis of the subsequent selection of specific remote sensing images.
[0091] Local area improvement: Based on the overlap range between the preset images, the final remote sensing image of the current target area is segmented into the overlap range area, and the remaining area except the overlap range area is used as the remaining supplementary area of the final remote sensing image. The corresponding route with the largest acquisition efficiency value CJX is selected by the drone to collect and supplement the remaining supplementary area of the current target area, and the remaining supplementary area image collected by the drone is fused with the final remote sensing image to generate a complete remote sensing image of the current target area.
[0092] The specific steps to obtain the collection efficiency value CJX of each UAV route are:
[0093] T1: Select a drone that meets the current collection mission requirements;
[0094] T2: Plan the UAV route based on the geographical location, area size and terrain characteristics of the remaining supplementary area, and obtain multiple planned routes for the current UAV to the remaining supplementary area;
[0095] It should be noted that it is necessary to ensure that the planned route is not within a no-fly zone or restricted flight area.
[0096] T3: Based on the planned multiple routes, calculate the total flight distance required for each route;
[0097] T4: Match the obtained total distance of each drone with the set distance reference range corresponding to the estimated flight time and the power consumption required for the flight; obtain the estimated flight time and the power consumption required for the flight of the drones corresponding to the remaining supplementary area;
[0098] T5: Collect wind speed and wind direction data of the current target area, calculate the angle between each flight route of the drone and the wind direction of the current target area, obtain the wind direction angle value of each route of the drone, and integrate the wind speed and corresponding wind direction angle value of each route into a meteorological influence set, and match each meteorological influence set obtained with multiple preset meteorological influence sets; set each preset meteorological influence set to correspond to a meteorological additional value; obtain the meteorological additional value of the drone on each route;
[0099] It should be noted that the maximum angle between each flight route of the UAV and the wind direction of the current target area is used as the wind direction angle value of each flight route of the UAV.
[0100] T6: Match the meteorological added value of each current route with the meteorological added value reference range corresponding to the set required additional time and required power consumption; obtain the required additional time and required power consumption of each current route, add the estimated flight time and required flight power consumption of each route in step T4 to the corresponding additional time and additional power consumption, and obtain the total flight time and total flight power of the drone on each route;
[0101] T7: Based on the comparison results of the total flight power of each route and the corresponding total power capacity of the drone, the required charging times of the drone on each route are obtained;
[0102] The total distance of the target route, the total flight time, and the number of required charging times are multiplied by the corresponding preset weight coefficients and then summed to obtain a result 1. The collection efficiency value of the target route drone is obtained by dividing the integer 1 by the result 1;
[0103] T8: Select the corresponding route with the largest collection efficiency value as the flight route for the current drone to collect the remaining supplementary area.
[0104] It should be noted that the route planning is based on the geographical location, area size and terrain characteristics, and a comprehensive analysis is conducted on the total distance of each route, the total flight time and the number of times required to charge, so as to comprehensively evaluate the collection efficiency of the UAV on the corresponding route. At the same time, by collecting wind speed and wind direction data, calculating the wind direction angle, and integrating the meteorological influence set, the influence of wind direction and wind speed on the UAV collection work in each route is considered, thereby improving the accuracy of data analysis.
[0105] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.
Claims
1. A remote sensing image feature recognition method, characterized in that: include: Obtain remote sensing images: Register and log in to the corresponding account to access different satellite data provision platforms, download geological remote sensing images of the target area; and perform preliminary screening of the downloaded remote sensing images based on the comparison results of the numerical evaluation values of each satellite data provision platform, and use the remote sensing images after preliminary screening as the preset images; According to the comparison results of the numerical quality evaluation values of each satellite data provider platform, the downloaded remote sensing images are initially screened. The specific screening steps are as follows: The spatial resolution, spectral resolution and temporal resolution of each satellite data provider platform are obtained; the spatial resolution, spectral resolution and temporal resolution of the corresponding satellite data provider platform are multiplied by the corresponding preset weight coefficients and then summed to obtain the numerical quality evaluation value of the corresponding satellite data provider platform; a threshold value of the numerical quality evaluation value is set, and the numerical quality evaluation value of each satellite data provider platform is compared with the set threshold value. If the numerical quality evaluation value of the corresponding satellite data provider platform is greater than the set threshold value, the geological remote sensing image downloaded by the platform is used as the preset image; Aerial photography data acquisition: register the selected preset images, identify and extract the overlap range of the preset images, and randomly select a set range from the obtained overlap range as a reference local image; extract the reference local image corresponding to the selected range of the current preset images; then use the drone to take aerial photos of the selected set range to obtain the drone image data of the selected set range; Image similarity analysis: convert the UAV image data and the corresponding reference local image into GeoTIFF format; preprocess the converted GeoTIFF images; after preprocessing, analyze the visual similarity between the UAV image data and the corresponding reference local image to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image; Geomorphic similarity analysis: extract the digital elevation model DEM from the reference local image corresponding to the current target area, recorded as the reference local image DEM, and extract the digital elevation model DEM from the drone image data, recorded as the drone DEM; pre-process the extracted DEM data; analyze the geomorphic similarity between the drone image data and the corresponding reference local image, and obtain the geomorphic similarity evaluation index SUG of the current target area corresponding to the reference local image; Select remote sensing images: Based on the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG of the reference local image corresponding to the current target area; according to the formula Perform normalized calculation to obtain the selected evaluation value XQP of the reference local image corresponding to the current target area; where gv1 and gv2 are the influencing weight factors of the visual similarity evaluation index SJP and the topographic similarity evaluation index SUG respectively; Arrange the selected evaluation values XQP corresponding to each reference local image of the current target area in descending order, and select the reference local image with the largest evaluation value XQP as the marked reference local image; extract the preset image corresponding to the marked reference local image, and use the preset image as the final remote sensing image for feature recognition of the current target area; Local area improvement: Based on the overlapping range between the preset images, the final remote sensing image of the current target area is segmented into the overlapping range area, and the remaining area except the overlapping range area is used as the remaining supplementary area of the final remote sensing image. The corresponding route with the largest collection efficiency is selected by the drone to collect and supplement the remaining supplementary area of the current target area, and the remaining supplementary area image collected by the drone is fused with the final remote sensing image to generate a complete remote sensing image of the current target area.
2. A remote sensing image feature recognition method according to claim 1, characterized in that: The visual similarity between the drone image data and the corresponding reference local image is analyzed. The specific analysis steps are as follows: Extract reflectance values from each band of the drone image data of the current target area and the corresponding reference local image; calculate the reflectance difference for each pixel point of the drone image data and the corresponding reference local image, that is, calculate by the formula reflectance difference value (i, j) = WX (i, j) - WR (i, j), and take the absolute value of the calculated result; Where i and j represent the row and column coordinates of the pixel, respectively, WX represents the reflectivity value of the corresponding reference local image, and WR represents the reflectivity value of the drone image data; Extract each group of reflection difference values of the reference local image corresponding to the current target area, calculate the mean, and use the calculated mean as the difference mean of the reference local image corresponding to the current target area; calculate each group of reflection difference values of the reference local image corresponding to the current target area using the standard deviation formula, and use the calculated result as the difference dispersion value of the reference local image corresponding to the current target area; extract the maximum reflection difference value and the minimum reflection difference value from each group of reflection difference values of the reference local image corresponding to the current target area, calculate the difference, and use the calculated difference as the maximum difference of the reference local image corresponding to the current target area; The difference mean, difference dispersion and difference maximum values corresponding to the reference local image of the current target area are multiplied by the corresponding preset weight coefficients and then summed to obtain the difference evaluation index CY1 of the reference local image of the current target area.
3. A remote sensing image feature recognition method according to claim 2, characterized in that: The specific steps to obtain the visual similarity evaluation index SJP of the reference local image corresponding to the current target area are: The reference local image and drone image data corresponding to the current target area are processed through the GLCM tool in the image processing software, and the same GLCM parameters are set; the frequency distribution of each pixel pair of the reference local image and drone image data corresponding to the current target area is calculated respectively to generate a GLCM matrix; the contrast, homogeneity, energy and entropy parameters are extracted from each generated GLCM matrix; the ratio of the reference local image corresponding to the current target area and the drone image data in terms of contrast, homogeneity, energy and entropy parameters is calculated respectively, and the difference between the ratio calculated for each parameter and the integer 1 is calculated, and the absolute value is taken; the mean of each group of differences after taking the absolute value of the reference local image corresponding to the current target area is calculated, and the mean is used as the texture evaluation index CY2 of the reference local image corresponding to the current target area; Identify and extract the edge information of the reference local image and drone image data corresponding to the current target area, and extract the contour from the edge information; for the two extracted contours, calculate the area corresponding to the two contours; calculate the Hu moment for the two extracted contours, and divide the Hu moment eigenvalues corresponding to the two contours by the corresponding contour area to obtain the 7 normalized Hu moments corresponding to the reference local image and drone image data of the current target area; calculate the ratio of the 7 normalized Hu moments of the reference local image corresponding to the current target area to the 7 normalized Hu moments of the drone image data one by one; calculate the difference between the ratio calculated by each parameter and the integer 1, and take the absolute value; calculate the mean of each group of differences after taking the absolute value of the reference local image corresponding to the current target area, and use the mean as the contour evaluation index CY3 of the reference local image corresponding to the current target area; Substitute the difference evaluation index CY1, texture evaluation index CY2 and contour evaluation index CY3 obtained by analyzing the reference local image corresponding to the current target area into the formula Normalized calculation is performed to obtain the visual similarity evaluation index SJP of the current target area corresponding to the reference local image; where CY1 参考 CY2 参考 and CY3 参考 are the set reference difference evaluation index, reference texture evaluation index and reference contour evaluation index respectively; zy1, zy2 and zy3 are the influencing weight factors of the difference evaluation index CY1, the texture evaluation index CY2 and the contour evaluation index CY3 respectively.
4. A remote sensing image feature recognition method according to claim 3, characterized in that: The similarity between the UAV image data and the corresponding reference local image is analyzed. The specific analysis steps are as follows: Extract corresponding terrain characteristic values from the drone image DEM and the corresponding reference local image DEM; the terrain characteristic values include slope, aspect, curvature and altitude; The slope characteristic value of each pixel point of the reference local image DEM corresponding to the current target area and the slope characteristic value of the pixel point of the drone image DEM are calculated to obtain the slope similarity value PM of the reference local image DEM and the drone image DEM corresponding to the current target area; that is, through the formula Calculated; where Sft and Sut represent the slope characteristic values of the drone image DEM and the corresponding reference local image DEM at pixel t, respectively; where t = 1, 2 or E, and E is the total number of pixels; The same steps as above are used to calculate the aspect similarity value, curvature similarity value and altitude similarity value of the reference local image DEM and the drone image DEM at each pixel point in the current target area, and they are marked as PN, PC and PZ respectively.
5. A remote sensing image feature recognition method according to claim 4, characterized in that: The specific steps to obtain the reference local image topography similarity assessment index SUG corresponding to the current target area are: According to the formula Normalization processing is performed to obtain the topographic similarity assessment index SUG of the current target area corresponding to the reference local image; PM 参考 PN 参考 、PC 参考 and PZ 参考 are the set reference slope similarity value, reference aspect similarity value, reference curvature similarity value and reference altitude similarity value respectively; ce1, ce2, ce3 and ce4 are the influencing weight factors of slope similarity value PM, aspect similarity value PN, curvature similarity value PC and altitude similarity value PZ respectively.
6. A remote sensing image feature recognition method according to claim 5, characterized in that: The specific steps to obtain the collection efficiency value CJX of each UAV route are: T1: Select a drone that meets the current collection mission requirements; T2: Plan the UAV route based on the geographical location, area size and terrain characteristics of the remaining supplementary area, and obtain multiple planned routes for the current UAV to the remaining supplementary area; T3: Based on the planned multiple routes, calculate the total flight distance required for each route; T4: Match the total distance of each drone obtained with the set distance reference range corresponding to the estimated flight time and the power consumption required for the flight; obtain the estimated flight time and power consumption required for the drone corresponding to the remaining supplementary area; T5: Collect wind speed and wind direction data of the current target area, calculate the angle between each flight route of the drone and the wind direction of the current target area, obtain the wind direction angle value of each route of the drone, and integrate the wind speed and corresponding wind direction angle value of each route into a meteorological influence set, and match each meteorological influence set obtained with each preset meteorological influence set; set each preset meteorological influence set to correspond to a meteorological additional value; obtain the meteorological additional value of the drone on each route; T6: Match the meteorological added value of each current route with the meteorological added value reference range corresponding to the set required additional time and required power consumption; obtain the required additional time and required power consumption of each current route, add the estimated flight time and required flight power consumption of each route in step T4 to the corresponding additional time and additional power consumption, and obtain the total flight time and total flight power of the drone on each route; T7: Based on the comparison results of the total flight power of each route and the corresponding total power capacity of the drone, the required charging times of the drone on each route are obtained; The total distance of the target route, the total flight time, and the number of required charges are multiplied by the corresponding preset weight coefficients and then summed to obtain result one. The collection efficiency value of the target route drone is obtained by dividing the integer 1 by result one.
