Method and system for environmental monitoring based on images taken by a drone

Abstract

The present application belongs to the technical field of computer vision and environment monitoring, and particularly relates to an environment monitoring method and system based on images taken by a UAV, which comprises the following steps: obtaining images of a stockyard collected by a UAV and performing superpixel segmentation; extracting color space statistical features, gray level co-occurrence matrix features and gradient features of each superpixel block; constructing a light saturation decoupling relationship formula, and removing shadow interference by using saturation difference; constructing a texture immersion smoothing relationship formula, and removing dark ore interference by using texture energy and gradient variance; calculating an environment compliance index based on a light saturation decoupling index and a texture immersion smoothing index, generating a heat map, and controlling dust suppression equipment according to the heat map. The present application effectively reduces misjudgment caused by backlight shadow and ore color by decoupling physical optical features, realizes accurate identification of the wet state of the surface of the stockyard, and can save water resources and improve dust suppression effect.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and environmental monitoring technology. More specifically, this invention relates to environmental monitoring methods and systems based on images captured by unmanned aerial vehicles (UAVs). Background Technology

[0002] In the storage and transshipment of bulk commodities such as iron ore at ports, regular dust suppression spraying operations are required on the surface of iron ore storage yards to meet environmental dust pollution control standards. The implementation and effectiveness control of these spraying operations directly determine the effectiveness of dust control and the rationality of water resource utilization. Currently, monitoring the dust suppression effect of spraying mainly relies on manual inspections or simple video monitoring. These traditional monitoring methods suffer from low operational efficiency, limited coverage, and inability to provide real-time monitoring across the entire area. They not only fail to provide accurate data support but are also susceptible to human error, leading to biased judgments and failing to provide reliable decision-making basis for spraying operations.

[0003] Existing technologies also employ automated monitoring schemes based on image pixel brightness threshold segmentation. The logic behind this is that wet surfaces are relatively less bright than dry surfaces; by comparing image pixel brightness values ​​with a preset threshold, areas with brightness below the threshold are identified as wet. However, this method based on a single brightness threshold has physical logic flaws when applied to real-world iron ore storage yard scenarios, easily leading to distorted monitoring results.

[0004] On the one hand, the surface of iron ore stockpiles is naturally uneven, and under natural light, a large number of shadow areas will form on the shaded side of the stockpiles. The pixel brightness value of such areas is extremely low, and existing technology cannot effectively distinguish between dry shadow areas and actual wet dark areas. It is easy to misjudge dry shadows as wet areas, which will lead to missed spraying and create dust hazards. On the other hand, the optical properties of iron ore from different producing areas vary greatly. For example, magnetite is black and shiny, while hematite is reddish-brown. The color of the ore itself varies greatly. Existing technology is prone to misjudging dry dark magnetite as wet areas. If the judgment threshold is raised to adapt to dark ore, it will cause excessive spraying of light-colored ore areas, which will not only waste water resources, but also cause the moisture content of the ore to exceed the standard, directly affecting the accuracy of the weight of ore trade settlement.

[0005] In summary, the existing monitoring methods based on brightness thresholds cannot overcome the dual interference of light and shadow and the natural color of the ore, making it difficult to accurately identify the surface wetness of iron ore stockpiles. This leads to problems such as inaccurate spray control and poor dust suppression. Therefore, the industry urgently needs an intelligent monitoring method and system that can effectively eliminate the interference of shadow and the natural color of the ore and has high monitoring accuracy. Summary of the Invention

[0006] To address the technical problem that existing monitoring methods based on image pixel brightness threshold segmentation cannot effectively distinguish between shaded and moist areas in stockpiles, or between dry dark ore and moist ore, leading to inaccurate spray control, poor dust suppression, and water waste, this invention proposes an environmental monitoring method and system based on images captured by drones.

[0007] In a first aspect, the present invention provides an environmental monitoring method based on images captured by a drone, comprising: acquiring orthophotos of a target area collected by the drone, and dividing the orthophotos into multiple physically meaningful superpixel blocks; extracting color space features, texture features, and gradient features of each superpixel block, wherein the color space features include brightness, saturation, and neighborhood average saturation, and the texture features include contrast and energy; calculating a light saturation decoupling index characterizing the humidity level of the area and eliminating shadow interference based on the saturation, brightness, and neighborhood average saturation; calculating a texture wetting smoothness index characterizing the surface smoothness of the area and eliminating dark mineral interference based on the energy, contrast, and gradient features; performing nonlinear compression on the light saturation decoupling index and weighted fusion with the texture wetting smoothness index to calculate an environmental compliance index, and determining the humidity state of the target area based on the environmental compliance index to control the operation of dust suppression equipment; wherein the formula for calculating the environmental compliance index is:

[0008] ;

[0009] In the formula, Indicates the first Environmental compliance index of each superpixel block The light saturation decoupling index. The texture wetting smoothness index. and These are the first weighting coefficient and the second weighting coefficient, respectively. This represents the natural logarithm function.

[0010] This invention first acquires orthophotos of the stockpile using a drone and performs superpixel segmentation, extracting multi-dimensional features such as brightness, saturation, texture energy, contrast, and gradient variance for each sub-region. Then, it constructs a light saturation decoupling index, utilizing the physical characteristic of decreased brightness but increased saturation in wet areas to eliminate shadow interference. Simultaneously, it constructs a texture wetting smoothing index, based on the water film filling effect leading to increased texture energy and decreased contrast and gradient variance to distinguish between dark, dry ores and wet ores. Finally, the two indices are weighted and fused into an environmental compliance index, generating a heatmap. This method physically distinguishes between illuminated black and physical black, effectively reducing misjudgments caused by backlighting shadows and the natural color of the ore, achieving accurate identification of the surface wetness of the stockpile. This guides dust suppression equipment to spray as needed, saving water resources and improving dust suppression effectiveness.

[0011] Preferably, the step of acquiring orthophotos of the target area by a drone and dividing the orthophotos into multiple physically meaningful superpixel blocks includes: controlling a drone equipped with a high-resolution visible light camera to acquire orthophotos of an iron ore stockpile at a preset altitude; converting the acquired orthophotos from the RGB color space to the CIELAB color space; and using a simple linear iterative clustering algorithm to divide the converted image into a preset number of superpixel blocks, generating a superpixel label matrix containing the label of each pixel.

[0012] Preferably, the color space features, texture features, and gradient features of each superpixel block are extracted, including: converting all pixels within the superpixel block in the RGB color space to the HSV color space, calculating the normalized average brightness and average saturation, and calculating the mean of the average saturation of all adjacent superpixel blocks as the neighborhood average saturation; converting the superpixel block in the RGB color space into a grayscale image and compressing the grayscale levels, constructing a grayscale co-occurrence matrix in multiple directions, and calculating the mean of contrast and energy; using the Sobel operator to calculate the gradient magnitude of all pixels in the grayscale image, and calculating the normalized gradient variance.

[0013] Preferably, the formula for calculating the light saturation decoupling index, which characterizes the humidity level of a region and eliminates shadow interference, is as follows:

[0014]

[0015] In the formula, Indicates the first The illumination saturation decoupling index of each superpixel block Indicates the first Average saturation of each superpixel block Indicates the first The average brightness of each superpixel block Indicates local saturation gain. Indicates the brightness suppression coefficient. This represents the difference amplification factor. Represented by natural constant An exponential function with base 0.

[0016] By introducing local saturation gain and an exponential function, and taking advantage of the optical property that shadows do not change saturation, it is possible to distinguish between low brightness caused by shadows and low brightness caused by moisture, thus solving the problem of misjudgment in shadow areas.

[0017] Preferably, the value of the local saturation gain is determined by the following method: calculating the difference between the average saturation of the current superpixel block and the average saturation of the neighborhood of the superpixel block; determining whether the difference is greater than zero; if the difference is greater than zero, then the difference is determined as the value of the local saturation gain; if the difference is less than or equal to zero, then zero is determined as the value of the local saturation gain.

[0018] Preferably, the formula for calculating the texture wetting smoothness index, which characterizes the surface smoothness of the region and eliminates interference from dark minerals, is as follows:

[0019]

[0020] In the formula, Indicates the first Texture wetting smoothness index of each superpixel block Indicates the first Gray-level co-occurrence matrix energy of each superpixel block Indicates the first Gray-level co-occurrence matrix contrast of each superpixel block Indicates the first The gradient variance of each superpixel block Indicates the energy-weighted power exponent. Indicates the flash suppression coefficient. To prevent extremely small constants with a denominator of zero.

[0021] Based on the energy, contrast, and gradient variance of texture, combined with the physical properties of dry surface roughness and water film filling effect, it is possible to effectively distinguish between natural black and shiny magnetite and actual wet ore, thereby avoiding misjudgment of dark-colored minerals.

[0022] Preferably, the step of determining the humidity status of the target area based on the environmental compliance index to control the operation of the dust suppression equipment includes: mapping the environmental compliance index of all superpixel blocks back to the original image space to generate a heat map; setting an environmental compliance threshold; if the environmental compliance index of a certain area is greater than the environmental compliance threshold, then the area is determined to be effectively humid; if the environmental compliance index of a certain area is less than or equal to the environmental compliance threshold, then the area is determined to need water replenishment, and a spraying operation command is generated based on the location coordinates of the area and sent to the dust suppression equipment.

[0023] This solution directly converts image recognition results into equipment control signals, realizing automated closed-loop control from data acquisition and status determination to dust suppression operations. It ensures that water is precisely replenished only to dry areas, effectively avoiding water waste caused by indiscriminate spraying.

[0024] Preferably, the dust suppression equipment includes a dust suppression fog cannon or a water sprinkler truck, and the spraying operation instruction includes area data and location data of the area that needs to be watered.

[0025] Secondly, the present invention provides an environmental monitoring system based on images captured by a drone, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned environmental monitoring method based on images captured by a drone is implemented.

[0026] By adopting the above technical solution, the environmental monitoring method based on UAV images is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0027] The technical solution of the present invention has the following beneficial technical effects:

[0028] By acquiring orthophotos of iron ore stockpiles using drones, and then segmenting and aggregating them into physically meaningful local areas using SLIC superpixel technology, multi-dimensional features are extracted. A light saturation decoupling index is constructed to accurately eliminate shadow interference, and a texture wetting smoothing index is used to eliminate the influence of the natural color of dark-colored ore. Finally, an environmental compliance index is obtained through nonlinear fusion, enabling the determination of the stockpile's wetness status and the automated control of dust suppression equipment. This solution solves the problem of dry shadow areas and dry dark-colored ore being misclassified as wet areas in existing technologies, effectively avoiding dust hazards caused by missed spraying, while reducing water waste from excessive spraying and trade settlement discrepancies caused by excessive ore moisture content. It significantly improves the accuracy and efficiency of spray dust suppression operations, meeting the environmental management requirements of port iron ore stockpiles. Attached Figure Description

[0029] Figure 1 This is a flowchart of the environmental monitoring method based on images captured by a drone according to the present invention;

[0030] Figure 2 This is a schematic diagram of an orthophoto taken by a drone at the iron ore stockpile site.

[0031] Figure 3 It is a monitoring result image generated using existing technology based on the brightness threshold method;

[0032] Figure 4 It is a monitoring result image generated using the environmental monitoring method based on images captured by UAVs according to the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0034] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] This invention discloses an environmental monitoring method based on images captured by a drone, referring to... Figure 1 This includes the following steps:

[0036] S1. Acquire orthophotos of the target area collected by the UAV and divide the orthophotos into multiple physically meaningful superpixel blocks.

[0037] Specifically, an industrial-grade drone equipped with a high-resolution visible light camera was used to acquire orthophotos of the target area, the iron ore stockpile. The acquisition altitude was set at 30 to 50 meters to ensure a ground resolution of better than 1 cm / pixel, thereby achieving clear acquisition of the surface texture details of the ore. During the acquisition process, the ambient light intensity was recorded simultaneously, and the camera exposure parameters were adjusted in real time to avoid large-area overexposure in the images.

[0038] Since the surface of iron ore is composed of discrete particles, direct analysis of single pixels is easily affected by noise and lacks statistical significance at the texture level. Therefore, this step uses a simple linear iterative clustering algorithm to segment the original orthophoto image into physically meaningful superpixel blocks. The specific operation is as follows: To improve the accuracy of clustering segmentation boundaries, the original RGB format image is first converted to the CIELAB color space. This is because the color representation of this color space is more in line with the visual perception of the human eye, and it can better perform clustering based on color and brightness similarity. Then, the segmentation parameter N is set, and clustering calculation is performed based on the converted CIELAB image to segment the image into N superpixel blocks. Each superpixel block corresponds to a local area of ​​about 10 cm to 20 cm in size on the surface of the iron ore stockpile. Finally, the superpixel label matrix is ​​output. Each pixel (x, y) in this matrix corresponds to a label i, which is used to identify the superpixel block to which the pixel belongs in the original RGB image, thus providing a region index for subsequent feature extraction.

[0039] The superpixel segmentation operation described above aggregates discrete pixels into physically meaningful local regions, effectively reducing the noise impact of single-pixel analysis and laying a data foundation for subsequent extraction of stable statistical features.

[0040] S2. Extract the color space features, texture features, and gradient features of each superpixel block. The color space features include brightness, saturation, and neighborhood average saturation. The texture features include contrast and energy.

[0041] For each superpixel block ,in For the superpixel index, the value ranges from 1 to N. A feature vector representing its physical state is extracted. The feature vector includes color space features, texture features, and gradient features. The specific extraction methods for each feature are as follows:

[0042] First, color space conversion and statistics are performed: based on the aforementioned superpixel label matrix. Locate and extract RGB format superpixel blocks of the corresponding region in the original orthophoto. To independently analyze reflectivity and color purity, superpixel blocks... All pixels within the model are converted from the RGB color space to the HSV color space. The HSV color model decouples brightness and saturation into independent features, thus providing an independent and reliable data foundation for subsequent mathematical modeling to eliminate shadows and the inherent color interference of minerals. After the conversion, the average brightness is calculated separately. The values ​​are normalized to the [0,1] interval, are dimensionless, and are used to characterize the intensity of reflected light in the superpixel region; average saturation. The values ​​are normalized to the [0,1] interval, are dimensionless, and are used to characterize the vividness of the color in the superpixel block region; neighborhood average saturation. Search for superpixel blocks The average saturation of all adjacent superpixel blocks is calculated and used as the parameter value. The HSV color space is a color model oriented towards human visual perception. It stands for Hue, Saturation, and Luminance. It decouples luminance and saturation into independent features, accurately representing the reflectivity and color purity of mineral areas. This provides independent and reliable color feature data for subsequent mathematical model construction to eliminate shadows and the inherent color interference of the mineral.

[0043] Next, gray-level co-occurrence matrix feature extraction is performed: In order to eliminate the interference of color information on surface microstructure analysis and to purely extract texture features, the feature extraction is based on the label matrix. Locating and acquiring RGB format superpixel blocks The image is converted to grayscale and compressed to 64 levels to reduce subsequent computation. In the specific calculation process, the superpixel label matrix is ​​first used... Obtain target superpixel block Find the minimum bounding rectangle region and extract the corresponding grayscale sub-image; then use the label matrix As a mask, when statistically analyzing the distribution of pixel pairs in the 0°, 45°, 90°, and 135° directions, only when the step size is... Both pixels within a certain distance are located inside the mask (i.e., their corresponding label values ​​are both equal to 1). Only when a pixel crosses a boundary is it included in the frequency statistics, thus effectively eliminating invalid pixel pairs that interfere with the data. Finally, the mean of the texture features in four directions is calculated to obtain two core physical texture indicators: contrast ratio and other parameters. It reflects the depth and clarity of the grooves and textures on the surface of the ore; energy This reflects the consistency and uniformity of the surface texture of the ore.

[0044] Finally, gradient feature extraction is performed: the Sobel operator is used to calculate the gradient magnitude of all pixels in the grayscale image after the grayscale levels are compressed, and the gradient variance is calculated based on the gradient magnitude. All gradient feature values ​​are normalized to the [0,1] interval to detect strong specular reflection points on the ore surface. The Sobel operator, a classic 3×3 convolution operator used in digital image processing to calculate pixel grayscale gradient magnitudes and detect image edges, is used in this scheme to calculate the gradient magnitude of pixels within a superpixel block and solve for the gradient variance, thereby detecting strong specular reflection points on the ore surface and distinguishing the surface texture features of dry and wet ores.

[0045] Through the above multi-dimensional feature extraction, we can comprehensively obtain physical information that reflects the lighting conditions, color purity, and surface microstructure of the target area, providing complete data support for subsequent decoupling analysis of shadows and the inherent color interference of minerals.

[0046] S3. Calculate the light saturation decoupling index, which characterizes the humidity of the region and eliminates shadow interference, based on saturation, brightness, and neighborhood average saturation.

[0047] To address the technical problem in existing technologies where dry shadow areas are easily misidentified as wet areas, this step constructs a mathematical model based on physical principles to distinguish between shadow black and wet black. The physical logic is as follows: when the mineral surface is wet, light entering the water film undergoes total internal reflection, leading to a decrease in brightness (V). However, the water film enhances the color rendering of the mineral's true color, increasing saturation (S). In contrast, shadow areas on the mineral surface only experience a decrease in brightness (V) due to insufficient light. Lacking direct light excitation and influenced by scattered skylight, saturation (S) does not increase and may even be lower than in the surrounding illuminated areas. This embodiment utilizes the difference between the color rendering effect of water film reflection and the color distortion characteristics of shadows caused by Rayleigh scattering.

[0048] Based on the above physical logic, a light saturation decoupling index is constructed to characterize the regional humidity level and eliminate shadow interference. The calculation formula is as follows:

[0049]

[0050] In the formula, Indicates the first Average saturation of each superpixel block Indicates the first Average brightness of each superpixel block; This is the brightness suppression coefficient. Both are dimensionless parameters, and are used as difference amplification factors. In this embodiment, The value is 2. The value is 5; The local saturation gain is calculated using the following formula: , The local saturation gain is 0 when the average saturation of the current superpixel block is lower than the average saturation of its surrounding neighborhood.

[0051] It is important to note that the light saturation decoupling index constructed in this invention is an empirical physical model derived from the natural laws of physical optics and fitted with a large amount of measured data. In existing image processing techniques, surface wetness is typically determined solely by the reduction in brightness characteristics, but this cannot avoid misjudgments caused by natural shadows. The natural law upon which the relationship in this invention is based is the total internal reflection effect of water films: when the ore surface is wetted by a water film, light undergoes total internal reflection within the water film, reducing diffuse reflection, which leads to a significant increase in color saturation captured by industrial cameras; while dry shadow areas are mainly affected by Rayleigh scattering from the sky, and their saturation does not increase. To characterize this physical difference in computer vision algorithms, this invention derives the above relationship by performing multinomial regression fitting on massive sample data collected from the stockpile site. This relationship utilizes a natural exponential function. Its nonlinear smoothing properties enable nonlinear amplification of wet features and effective suppression of shadow features.

[0052] and The acquisition logic is as follows: First, sample images of the stockpile at different times (early morning, noon, and dusk) are collected, and the feature distribution of typical shaded and wet areas is extracted; As a brightness suppression coefficient, its value is positively correlated with the ambient light intensity. Linear regression analysis on the sample set ensures that under strong light conditions (brightness... When the mean value is high, it can suppress the interference of brightness on saturation to a greater extent; As a difference amplification factor, it is used to calculate the difference between dry shadows and real wet superpixel blocks in the sample set. The Fisher criterion score on the index is selected as the calibration value that maximizes the inter-class distance between the two classes of samples. When the ambient light sensor detects that the ambient light intensity exceeds a preset threshold, the system automatically... Adjust the settings by 10%-20% simultaneously to maintain recognition stability.

[0053] To verify the rationality of the above calculation formula, a calculation demonstration was conducted using typical engineering data. The scenario included a dry, shaded area (Scenario A) and a realistic wet area (Scenario B). The specific calculation process and results are as follows:

[0054] Scenario A: This area is in shadow, resulting in low brightness due to insufficient lighting. Normalized brightness The value is 0.2; meanwhile, due to the absence of a water film to enhance color rendering, its saturation is low and does not exceed the normalized saturation of the surrounding environment. The value is 0.1; the saturation of the surrounding neighborhood is... It is 0.15;

[0055] Calculate the local saturation gain: due to Take the maximum value ,Right now ;

[0056] Calculation of the light saturation decoupling index: Exponent term: ; items in brackets: ;

[0057] The final result is: Therefore, the light saturation decoupling index of the dry shadow region is 0, indicating that the region is dry, which is consistent with the actual physical state and achieves effective elimination of shadow interference.

[0058] Scenario B: This area is humid, and the brightness is low due to the absorption of light by the water film. Normalized brightness The value is 0.2; at the same time, the water film makes the ore's natural color more vibrant and its saturation higher, with a normalized saturation of 0.2. The saturation level is 0.6 in the surrounding dry area. It is 0.2;

[0059] Calculate the local saturation gain: ; Index term: ; items in brackets: ;

[0060] The final result is: Therefore, it can be seen that the light saturation decoupling index of a truly humid region is significantly greater than 0, and the system can accurately identify the humidity characteristics of that region.

[0061] This step, based on the difference in physical properties between surface wetness and shadow on the ore, uses a mathematical model to successfully eliminate the interference of shadow on the monitoring results of wetness status, thus avoiding the situation where dry shadow areas are misjudged as wet areas.

[0062] S4. Based on energy, contrast and gradient characteristics, calculate the texture wetting smoothness index, which characterizes the surface smoothness of the region and eliminates interference from dark minerals.

[0063] To address the technical problem in existing technologies where dry, dark-colored ores are easily misidentified as wet areas, this step utilizes the water film leveling effect to construct a texture discrimination model to distinguish between dark, dry ores and wet ores. The physical logic is as follows: although dry, dark-colored ores are inherently black, their surface has numerous gaps due to particle dispersion, resulting in a rough surface and a high contrast C value. Furthermore, the cleavage planes of the ore crystals contribute to the gradient variance. The contrast value is high; however, regardless of the base color, the surface water film of moist minerals fills the gaps between particles to form a smooth surface, resulting in a low contrast value (C) and a high energy value (E). Furthermore, the continuous reflection of the water film leads to a gradient variance. The value is small.

[0064] Based on the above physical logic, a texture wetting smoothness index is constructed to characterize the surface smoothness of a region while eliminating interference from dark minerals. The calculation formula is as follows:

[0065]

[0066] In the formula, It is an energy-weighted power exponent. Both are dimensionless parameters, namely the flash suppression coefficient and the flash suppression coefficient. In this embodiment, The value is 1.5. The value is 10; For the first Gray-level co-occurrence matrix contrast of each superpixel block Indicates the first Gradient variance of each superpixel block; Representing the The gray-level co-occurrence matrix energy of each superpixel block is a texture feature used to characterize the texture consistency and uniformity of the ore surface. Its value is normalized to the [0,1] interval. To prevent extremely small constants with a denominator of zero, the value is taken as... Among them, all input texture feature values All values ​​are normalized. This is a flash suppression term, used to suppress interference from the cleavage planes of iron ore crystals on the determination of wettability.

[0067] The aforementioned texture wetting smoothing index essentially belongs to the multi-source micro-texture feature nonlinear fusion operator in the field of image feature processing. It extracts the energy of the gray-level co-occurrence matrix. Gray-level co-occurrence matrix contrast and gradient variance Water film filling is a well-known technique in digital image processing for characterizing the microstructure of physical surfaces. It is based on the water film filling effect: dry, dark-colored mineral surfaces are rough, exhibiting high contrast and large variance; however, when covered by a water film, the tiny pits are filled, forming a relatively smooth reflective surface. This is reflected in the underlying features of the image as the energy of the gray-level co-occurrence matrix. The contrast of the gray-level co-occurrence matrix increases sharply. and gradient variance Significant reduction. The energy features of the increased gray-level co-occurrence matrix are placed in the numerator, while the contrast and gradient variance features of the decreased gray-level co-occurrence matrix are placed in the denominator, thus amplifying the features of the smooth physical state. Furthermore, a power function is employed. The arithmetic square root is a well-known data scale compression and smoothing technique used in signal processing to prevent large variance interference caused by a single noise point (such as an individual highly reflective crystal surface).

[0068] and Results were obtained through calibration tests for different mineral types: Used to adjust sensitivity to energy distribution; for fine-grained minerals (such as pellets), this is achieved by reducing... To prevent misjudgment caused by overly uniform texture; The flare suppression coefficient is specifically set for minerals with highly reflective crystal cleavage planes (such as magnetite). During system initialization, the user inputs the mineral type of the current stockpile through the user interface, and the system automatically matches parameters based on a preset mineral characteristic library: for highly reflective minerals (such as magnetite), the following parameters are set: The value is set in the high range of 10-15 to forcibly eliminate gradient variance anomalies caused by crystal scintillation; for minerals with weak reflectivity (such as limonite), then... The value was lowered to the 3-5 range to retain necessary surface microtexture information.

[0069] To verify the rationality of the above calculation formula, a calculation demonstration was conducted using typical engineering data. The scenario included a dry, dark magnetite region (Scenario C) and a moist ore region (Scenario D). The specific calculation process and results are as follows:

[0070] Scenario C: Dry, dark magnetite with a rough surface and crystalline flashes; grayscale co-occurrence matrix energy. The contrast ratio is 0.1. The gradient variance is 0.9. It is 0.8;

[0071] The calculation shows that: Therefore, it can be seen that the texture wetting smoothness index value of the dry dark magnetite area is extremely low. The system determines that the area is dry and is not affected by the dark color of the ore, thus achieving effective removal of interference from the ore's natural color.

[0072] Scenario D: The ore surface is filled with a water film, and the gray-scale co-occurrence matrix energy... The contrast ratio is 0.8. The gradient variance is 0.1. It is 0.1;

[0073] The calculation shows that: This shows that the texture smoothness index of the moist ore area is significantly higher than that of the dry dark magnetite area, and there is a large difference between the two values. The system can accurately identify the moist characteristics of this area.

[0074] The texture discrimination model in this step calculates the difference in texture characteristics between dry and wet ore surfaces, effectively avoiding misjudgment of wetness caused by the darker color of the ore itself, and solving the technical problem of dry, dark-colored minerals being misjudged as wet areas.

[0075] S5. Calculate the environmental compliance index based on the light saturation decoupling index and the texture wetting smoothness index, and determine the humidity status of the target area based on the environmental compliance index to control the operation of the dust suppression equipment.

[0076] The aforementioned illumination saturation decoupling index and texture wetting smoothness index are nonlinearly fused to calculate the final environmental compliance index. This allows for a comprehensive determination of the humidity status of the target area, and the calculation formula is as follows:

[0077]

[0078] In the formula, The light saturation decoupling index; The texture wetting smoothness index; These are the first weighting coefficient and the second weighting coefficient, respectively. Both can be adaptively adjusted according to the site environment of the iron ore stockpile. In this embodiment, we take... . The natural logarithm function is used for smoothing. The numerical range of which makes it consistent with The dimensions are more compatible.

[0079] It should be noted that the environmental compliance index is a comprehensive trigger control signal that can directly drive the underlying industrial dust suppression equipment. Considering the differences in data volume for different physical characteristics, a logarithmic function is introduced into the formula. This is a well-known dynamic range compression technique in digital signal processing, used to nonlinearly compress decoupled feature values, preventing data overflow and ensuring that its dimensions and weights match those of the texture feature values ​​during fusion. By fusing the two types of features extracted based on physical laws, this index accurately characterizes the actual physical water shortage state of the local area of ​​the storage yard.

[0080] Weighting coefficient and The settings follow a logic of prioritizing optics and enhancing texture, derived through multi-factor regression analysis of calibration areas with known wet conditions. In actual operation, the system dynamically balances factors based on real-time environmental feedback: when drastic fluctuations in ambient light or large areas of dark minerals are detected, the system reduces... And improve simultaneously This is to enhance system stability, as physical textures are more stable than color features under extreme lighting conditions. The specific adjustment criterion is: if the average brightness of the current superpixel block... If the image is within the ideal imaging range of [0.4, 0.6], then maintain... ;like If the deviation from this range exceeds 30%, it will be automatically adjusted. The proportion of [something] is increased to over 0.6%, thereby achieving adaptive closed-loop monitoring for complex field environments.

[0081] After obtaining the environmental compliance index of all superpixel blocks, the monitoring data is post-processed, and the dust suppression equipment is automatically controlled based on the processing results. The specific operation is as follows:

[0082] All superpixel blocks The values ​​are mapped back to the image space of the original orthophoto to generate a pseudo-color heat map covering the entire iron ore stockpile, visually displaying the distribution of humidity across the entire area; environmental compliance thresholds are set. In this embodiment Take 0.5, if a certain region If so, the area is considered effectively moist; if If the area is dry, the area is determined to be dry. The area and location coordinates of the dry areas in the entire field are counted. Based on the statistical results, a spraying operation instruction is generated and sent to the dust suppression equipment terminals such as dust suppression fog cannons and water trucks to realize the automated closed-loop control of spraying dust suppression operations.

[0083] Through the above comprehensive index calculation and threshold determination, this invention realizes the output of monitoring data on the dust suppression effect of spraying in iron ore stockpiles and the automated closed-loop control of dust suppression equipment, effectively ensuring the accuracy and efficiency of spraying dust suppression operations.

[0084] Combined with original scene images captured by drones ( Figure 2 ), existing technology monitoring results images ( Figure 3 ) and the monitoring results image of the method of the present invention ( Figure 4 A comparison of the effects was conducted: The original scene included a dark magnetite area, a backlit shadow area, and a truly wet area; existing technologies would misjudge the dark magnetite area and the backlit shadow area as wet areas, displaying them as having high humidity, resulting in missed spraying of the corresponding areas; however, using the monitoring method of this invention, the dark magnetite area and the backlit shadow area were successfully removed (displayed as background color in the image), achieving accurate identification of the wet state.

[0085] This invention also discloses an environmental monitoring system based on images captured by a drone, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the environmental monitoring method based on images captured by a drone according to this invention.

[0086] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0087] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for environmental monitoring based on images taken by a drone, characterized in that, include: Acquire orthophotos of the target area collected by the drone and segment the orthophotos into multiple physically meaningful superpixel blocks; Extract the color space features, texture features, and gradient features of each superpixel block. The color space features include brightness, saturation, and neighborhood average saturation, while the texture features include contrast and energy. According to the saturation, the brightness and the neighborhood average saturation, a light saturation decoupling index is calculated, which represents the wetness degree of the region and eliminates the interference of shadows. The light saturation decoupling index of the first superpixel block is: , Indicates the first Average saturation and average brightness of each superpixel block Indicates local saturation gain. Indicates the brightness suppression coefficient. This represents the difference amplification factor. Represented by natural constant An exponential function with base 0; Based on energy, contrast, and gradient characteristics, a texture wetting smoothness index is calculated to characterize the surface smoothness of the region while eliminating interference from dark minerals; Texture wetting smoothness index of each superpixel block for: , Indicates the first Gray-level co-occurrence matrix energy and gray-level co-occurrence matrix contrast of each superpixel block. Indicates the first The gradient variance of each superpixel block Indicates the energy-weighted power exponent. Indicates the flash suppression coefficient. To prevent extremely small constants with a denominator of zero; After nonlinear compression of the illumination saturation decoupling index, it is weighted and fused with the texture wetting smoothness index to calculate the environmental compliance index. The system determines the humidity level of the target area based on the environmental compliance index in order to control the operation of dust suppression equipment, including mapping the environmental compliance index of all superpixel blocks back to the original image space to generate a heat map. An environmental compliance threshold is set. If the environmental compliance index of a certain area is greater than the environmental compliance threshold, the area is determined to be effectively humid. If the environmental compliance index of a certain area is less than or equal to the environmental compliance threshold, the area is determined to need water replenishment, and a spraying operation instruction is generated and sent to the dust suppression equipment based on the location coordinates of the area. No. Environmental compliance index of each superpixel block for: ； and These are the first weighting coefficient and the second weighting coefficient, respectively. This represents the natural logarithm function.

2. The environmental monitoring method based on UAV-captured images according to claim 1, characterized in that, The process of acquiring orthophotos of the target area by the UAV and segmenting the orthophotos into multiple physically meaningful superpixel blocks includes: Control a drone equipped with a high-resolution visible light camera to collect orthophotos of an iron ore stockpile at a preset altitude; The acquired orthophoto image is converted from the RGB color space to the CIELAB color space; The transformed image is segmented into a preset number of superpixel blocks using a simple linear iterative clustering algorithm, generating a superpixel label matrix containing the label of each pixel.

3. The environmental monitoring method based on UAV-captured images according to claim 1, characterized in that, Extract the color space features, texture features, and gradient features of each superpixel block, including: Convert all pixels within a superpixel block in the RGB color space to the HSV color space, calculate the normalized average brightness and average saturation, and calculate the mean of the average saturation of all neighboring superpixel blocks as the neighborhood average saturation. The superpixel blocks in the RGB color space are converted into grayscale images and the grayscale levels are compressed. A grayscale co-occurrence matrix in multiple directions is constructed, and the mean values ​​of contrast and energy are calculated. The gradient magnitude of all pixels in the grayscale image is calculated using the Sobel operator, and the normalized gradient variance is calculated.

4. The environmental monitoring method based on images captured by a UAV according to claim 1, characterized in that, The value of the local saturation gain is determined by calculating the difference between the average saturation of the current superpixel block and the average saturation of the neighborhood of the superpixel block. Determine whether the difference is greater than zero; If the difference is greater than zero, then the difference is determined as the value of the local saturation gain; If the difference is less than or equal to zero, then zero is determined as the value of the local saturation gain.

5. The environmental monitoring method based on UAV-captured images according to claim 1, characterized in that, The dust suppression equipment includes a dust suppression fog cannon or a water sprinkler truck, and the spraying operation instruction includes the area data and location data of the area that needs to be watered.

6. An environmental monitoring system based on images captured by unmanned aerial vehicles (UAVs), characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the environmental monitoring method based on images captured by a UAV according to any one of claims 1-5.

Images