Mist sprayer and method based on multi-modal data control

Abstract

The invention relates to the technical field of spraying, in particular to a multi-modal data control-based mist sprayer and method, the multi-modal data control-based mist sprayer comprises a first rotating seat, a second rotating seat, a spraying shell, a plurality of sprayers, a fan and an adjustable nozzle, the first rotating seat is rotationally arranged on a supporting seat, the second rotating seat is rotationally arranged on the first rotating seat, and the adjustable nozzle is arranged on the second rotating seat; the adjustable nozzle comprises a sliding conical sleeve, a control piece and a limiting ring, the sliding conical sleeve is arranged on one side of the spraying shell in a sliding mode, the control piece is connected with the sliding conical sleeve, the limiting ring is fixed to the sliding conical sleeve, and the control piece is connected with the control piece. And the sealing ring is matched with the sealing ring on the injection shell. And the spraying range can be conveniently adjusted, so that the use effect is better.

Description

Technical Field

[0001] This invention relates to the field of spray technology, and more particularly to a sprayer and method based on multimodal data control. Background Technology

[0002] Dry fog cooling robots are innovative environmental temperature regulation devices designed for outdoor or semi-open spaces to alleviate the discomfort caused by high temperatures. They utilize a special spray system to highly refine water into micron-sized particles, creating a so-called "dry fog." These tiny water particles rapidly evaporate in the air, absorbing surrounding heat and effectively lowering the ambient temperature without causing slippery surfaces or water accumulation.

[0003] This robot, equipped with a high-precision ultrasonic or high-pressure micro-spray system, highly refines filtered water into tiny water particles with diameters between 10 and 30 micrometers, forming a "dry fog" that is visible to the naked eye but extremely light to the touch. This dry fog, suspended in the air, has a very large specific surface area, enabling it to complete the evaporation process in a very short time. During evaporation, water molecules absorb a large amount of sensible heat from the surrounding air, achieving phase change cooling, thereby significantly reducing the local ambient temperature. This typically lowers the perceived temperature by 5 to 12 degrees Celsius, with a significant and uniform cooling effect.

[0004] Existing cooling robots often spray according to a pre-set program, targeting a specific area. They cannot effectively adjust the spray range based on the density of people on site or the ambient temperature, thus reducing the cooling effect. Summary of the Invention

[0005] The purpose of this invention is to provide a fogging device and method based on multimodal data control, which aims to facilitate the adjustment of the spray range, thereby improving the performance.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a fogging device based on multimodal data control, comprising a support base and a movable wheel, the movable wheel being disposed at the bottom of the support base, and further comprising a first rotating base, a second rotating base, a spray housing, multiple nozzles, a fan, and an adjustable nozzle. The first rotating base is rotatably disposed on the support base, the second rotating base is rotatably disposed on the first rotating base, the spray housing is fixed on the second rotating base, the multiple nozzles are disposed inside the spray housing, the fan is disposed on one side of the nozzles, and the adjustable nozzle comprises a sliding conical sleeve, a control element, and a limiting ring. The sliding conical sleeve is slidably disposed on one side of the spray housing, the control element is connected to the sliding conical sleeve, and the limiting ring is fixed on the sliding conical sleeve and matches a sealing ring on the spray housing.

[0007] The fogging device based on multimodal data control also includes a cooling structure, which is disposed inside the spray housing.

[0008] The cooling structure includes multiple semiconductor cooling chips, a sealing strip, a heat dissipation cavity, a partition, and a circulating fan. The multiple semiconductor cooling chips are disposed on one side of the nozzle for cooling the spray. The sealing strip is disposed between the semiconductor cooling chips and the spray housing. The heat dissipation cavity is disposed on the outside of the semiconductor cooling chips. The partition is fixed in the heat dissipation cavity. The circulating fan is disposed in the heat dissipation cavity.

[0009] The fog cannon based on multimodal data control also includes a control module, which includes a site photography unit, an area selection unit, a crowd density calculation unit, and a spray parameter generation unit. The site photography unit is used to capture site images; The area selection unit is used to calculate the sunlit area based on the site image; The crowd density calculation unit is used to calculate crowd density based on images of sunlit areas; The spray parameter generation unit is used to calculate the coordinates and range of the spray area based on the pedestrian density, and to generate spray control parameters based on the coordinates and range.

[0010] The region selection unit includes an image conversion subunit, a brightness threshold segmentation subunit, and a region generation subunit. The image conversion subunit is used to convert RGB color images into grayscale images and eliminate image noise; The brightness threshold segmentation subunit is used to calculate the global brightness mean and standard deviation based on the site image and set a dynamic threshold, and compare the pixel brightness value with the dynamic threshold to generate a mask. The region generation sub-unit is used to eliminate small noise points and fill the voids in the sunlight area, while retaining continuous bright areas as the sunlight-illuminated area.

[0011] The pedestrian density calculation unit includes a pedestrian detection subunit, a pedestrian tracking subunit, a pedestrian density calculation subunit, and a density grading subunit. The pedestrian detection subunit is used to perform pedestrian flow analysis in the sunlit area using YOLOv8 to obtain the bounding box and confidence score of each pedestrian. The pedestrian tracking subunit is used to output a unique ID and motion trajectory for each pedestrian based on the detection box and motion prediction associated with pedestrians in continuous frames. The crowd density calculation subunit is used to calculate crowd density by mapping pixel area to the real-world scale through camera calibration. The density grading subunit is used to perform fog control grading based on pedestrian flow density to obtain density levels.

[0012] The spray parameter generation unit includes a fitting calculation subunit, an angle calculation subunit, and an atomization parameter calculation subunit. The fitting calculation subunit is used to generate multiple circumcircles based on the sunlight-irradiated area and the maximum spray range, and to obtain the center coordinate array and corresponding radius of the multiple circumcircles, wherein the corresponding radius is less than or equal to the maximum spray range; The angle calculation subunit is used to output the pitch and azimuth angles of the rotating seat based on the transformation of the center coordinate array to the coordinate system of the spray nozzle rotating seat; The atomization parameter calculation subunit is used to match the atomization particle size and flow rate based on the corresponding radius and density level.

[0013] The control module further includes a correction unit, which is used to acquire wind data of the irradiated area and correct the spray control parameters based on wind speed and wind direction.

[0014] The correction unit includes a wind data acquisition subunit, a spray offset compensation subunit, and a mist particle size adjustment subunit. The wind data acquisition subunit is used to acquire wind speed and wind direction angle in real time through wind speed and wind direction sensors. The spray offset compensation is used to calculate the spray angle offset based on the wind speed value and wind direction angle, and to adjust the pointing angle of the rotating seat. The mist particle size adjustment subunit is used to match the corresponding atomization particle size based on the wind speed level in order to reduce drift loss.

[0015] Secondly, the present invention also provides a fogging method based on multimodal data control, comprising: Collect site photos; Calculate the sunlit area based on site images; Calculate pedestrian density based on images of sunlit areas; The coordinates and range of the spray area are calculated based on the pedestrian density, and spray control parameters are generated based on the coordinates and range.

[0016] This invention relates to a fogging device and method based on multimodal data control. The fogging device mainly includes a support base and movable wheels installed at the bottom of the support base. The movable wheels can adopt a metal hub structure covered with wear-resistant rubber and are equipped with a braking device to ensure stable parking during operation and flexible steering during movement. The support base, as the basic load-bearing structure of the entire device, is welded from high-strength steel and has good pressure resistance and corrosion resistance, adapting to complex and changing outdoor operating environments. A first rotating seat is provided above the support base. This first rotating seat is rotatably mounted on the support base via a horizontal rotary bearing or rotary drive mechanism, realizing 360-degree continuous rotation of the spraying device in the horizontal direction. A second rotating seat is further provided on the first rotating seat. The second rotating seat realizes up-and-down pitch movement relative to the first rotating seat through a pitch adjustment mechanism (such as a hydraulic cylinder, electric push rod, or gear rack structure driven by a servo motor), thereby achieving a wide range of adjustment of the spray angle in the vertical direction. The spray shell is fixedly installed on the second rotating seat and rotates and pitches synchronously with it. The spray nozzle housing houses multiple high-pressure nozzles arranged in an array. These nozzles are connected to an external water source or an internal water tank via water supply lines and are powered by a high-pressure water pump. The nozzles can be selected from spiral or fan-shaped nozzles with excellent atomization effects, capable of atomizing water or other liquid agents into micron-sized particles, increasing coverage area and adsorption efficiency. A high-powered fan is located on one side of the nozzle. Driven by a motor, the high-speed airflow further propels the atomized water mist from the nozzle to distant target areas, significantly improving range and diffusion uniformity. The fan and nozzles work together to create a "wind-assisted mist" effect, effectively enhancing dust suppression, cooling, or disinfection effects.

[0017] One of the key innovations of this invention lies in the provision of an adjustable nozzle structure for dynamically adjusting the cross-sectional area and airflow pattern of the outlet, thereby controlling the concentration and diffusion angle of the spray. This adjustable nozzle includes a sliding conical sleeve, a control component, and a limiting ring. The sliding conical sleeve is coaxially slidably disposed on one side of the outlet end of the spray housing, and its shape is a tapered or expanding cone structure. The effective diameter of the outlet can be changed by axial sliding. The control component (which can be an electric push rod, a lead screw and nut mechanism, or a pneumatic cylinder) is connected to the sliding conical sleeve, receiving commands from the control system to achieve precise displacement control of the sliding conical sleeve. The limiting ring is fixed to the outer wall of the sliding conical sleeve, forming a sliding seal with a sealing ring on the spray housing, ensuring structural stability during sliding, preventing airflow leakage, and improving wind energy utilization efficiency. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a structural diagram of the fog cannon based on multimodal data control according to the present invention.

[0020] Figure 2 This is a right-side structural diagram of the fog cannon based on multimodal data control according to the present invention.

[0021] Figure 3 This is a cross-sectional view of the fog cannon based on multimodal data control according to the present invention.

[0022] Figure 4 This is a longitudinal cross-sectional view of the fog cannon based on multimodal data control according to the present invention.

[0023] Figure 5 This is a structural diagram of the control module of the present invention.

[0024] Figure 6 This is a structural diagram of the region selection unit of the present invention.

[0025] Figure 7 This is a structural diagram of the crowd density calculation unit of the present invention.

[0026] Figure 8 This is a structural diagram of the spray parameter generation unit of the present invention.

[0027] Figure 9 This is a structural diagram of the correction unit of the present invention.

[0028] Figure 10 This is a flowchart of the fog spraying method based on multimodal data control of the present invention.

[0029] Support base 101, casters 102, first rotating base 103, second rotating base 104, spray housing 105, nozzle 106, fan 107, adjustable nozzle 108, sliding conical sleeve 109, control component 110, limit ring 111, semiconductor cooling chip 112, sealing strip 113, heat dissipation cavity 114, partition 115, circulating fan 116, site photography unit 117, area selection unit 118, crowd density calculation unit 119, spray parameter generation unit 1 20. Correction unit 121. Image conversion subunit 122. Brightness threshold segmentation subunit 123. Region generation subunit 124. Pedestrian detection subunit 125. Pedestrian tracking subunit 126. Pedestrian density calculation subunit 127. Density grading subunit 128. Fitting calculation subunit 129. Angle calculation subunit 130. Atomization parameter calculation subunit 131. Wind data acquisition subunit 132. Spray offset compensation subunit 133. Fog particle size adjustment subunit 134. Detailed Implementation

[0030] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0031] First Embodiment Please see Figures 1-9 This invention provides a fogging device based on multimodal data control, including a support base 101 and a moving wheel 102. The moving wheel 102 is disposed at the bottom of the support base 101. It also includes a first rotating base 103, a second rotating base 104, a spray housing 105, multiple nozzles 106, a fan 107, and an adjustable nozzle 108. The first rotating base 103 is rotatably disposed on the support base 101, and the second rotating base 104 is rotatably disposed on the first rotating base 103. The spray housing 105 is fixed to the first rotating base 101. On the second rotating base 104, a plurality of nozzles 106 are disposed inside the spray housing 105. The blower 107 is disposed on one side of the nozzles 106. The adjustable nozzle 108 includes a sliding conical sleeve 109, a control element 110, and a limiting ring 111. The sliding conical sleeve 109 is slidably disposed on one side of the spray housing 105. The control element 110 is connected to the sliding conical sleeve 109. The limiting ring 111 is fixed on the sliding conical sleeve 109 and matches the sealing ring on the spray housing 105.

[0032] In this embodiment, the sprayer mainly includes a support base 101 and casters 102 mounted on the bottom of the support base 101. The casters 102 can be made of wear-resistant rubber-coated metal hubs and equipped with a braking device to ensure stable parking during operation and flexible steering during movement. The support base 101, as the basic load-bearing structure of the entire device, is welded from high-strength steel and has good pressure resistance and corrosion resistance, making it adaptable to complex and changing outdoor operating environments. A first rotating seat 103 is provided above the support base 101. The first rotating seat 103 is rotatably mounted on the support base 101 via a horizontal slewing bearing or a slewing drive mechanism, enabling the spraying device to rotate continuously 360 degrees in the horizontal direction. A second rotating seat 104 is further provided on the first rotating seat 103. The second rotating seat 104 achieves up-and-down pitch movement relative to the first rotating seat 103 through a pitch adjustment mechanism (such as a hydraulic cylinder, electric push rod, or a gear and rack structure driven by a servo motor), thereby achieving a wide range of adjustment of the spray angle in the vertical direction. The spray housing 105 is fixedly mounted on the second rotating base 104, rotating and pitching synchronously with it. Inside the spray housing 105 are multiple high-pressure nozzles 106 arranged in an array. These nozzles 106 are connected to an external water source or an internal water tank via water supply pipes and are powered by a high-pressure water pump. The nozzles 106 can be spiral or fan-shaped nozzles with excellent atomization effect, capable of atomizing water or other liquid agents into micron-sized particles, improving coverage area and adsorption efficiency. A high-power fan 107 is located on one side of the nozzles 106. Driven by a motor, the fan 107 generates a high-speed airflow that further propels the atomized water mist from the nozzles 106 to a distant target area, significantly improving range and diffusion uniformity. The fan 107 and the nozzles 106 work together to create a "wind-assisted mist" effect, effectively enhancing dust suppression, cooling, or disinfection effects.

[0033] One of the key innovations of this invention lies in the provision of an adjustable nozzle 108 structure, used to dynamically adjust the cross-sectional area and airflow pattern of the outlet, thereby controlling the concentration and diffusion angle of the spray. The adjustable nozzle 108 includes a sliding conical sleeve 109, a control component 110, and a limiting ring 111. The sliding conical sleeve 109 is coaxially slidably disposed in the effective passage of the spray housing 105. The control component 110 (which can be an electric push rod, a lead screw and nut mechanism, or a pneumatic cylinder) is connected to the sliding conical sleeve 109, receiving commands from the control system to achieve precise displacement control of the sliding conical sleeve 109. The limiting ring 111 is fixed to the outer wall of the sliding conical sleeve 109, forming a sliding seal with the sealing ring provided on the spray housing 105, ensuring structural stability during sliding and preventing airflow leakage, thus improving wind energy utilization efficiency.

[0034] The fogging device based on multimodal data control also includes a cooling structure, which is disposed within the spray housing 105.

[0035] The cooling structure includes multiple semiconductor cooling chips 112, a sealing strip 113, a heat dissipation cavity 114, a partition 115, and a circulating fan 116. The multiple semiconductor cooling chips 112 are disposed on one side of the nozzle 106 for cooling the spray. The sealing strip 113 is disposed between the semiconductor cooling chips 112 and the spray housing 105. The heat dissipation cavity 114 is disposed on the outside of the semiconductor cooling chips 112. The partition 115 is fixed inside the heat dissipation cavity 114. The circulating fan 116 is disposed inside the heat dissipation cavity 114.

[0036] The cooling structure includes multiple thermoelectric cooling chips 112 (also known as thermoelectric cooling chips or TEC modules), a sealing strip 113, a heat dissipation cavity 114, a partition 115, and a circulating fan 116. The multiple thermoelectric cooling chips 112 are arranged in an array on one side of the nozzle 106, preferably in the airflow channel area between the nozzle 106 and the fan 107, ensuring that the atomized water mist fully contacts the cold end surface of the cooling chips when passing through this area, achieving rapid heat exchange. Each thermoelectric cooling chip 112 has clearly defined hot and cold sides: the cold end faces the spray channel to absorb heat from the droplets and surrounding air, achieving cooling; the hot end extends into the outer heat dissipation cavity 114 to dissipate heat promptly.

[0037] To prevent cold air leakage and the infiltration of high-temperature external air from affecting cooling efficiency, an elastic, heat-resistant sealing strip 113 is provided between the edge of each semiconductor cooling chip 112 and the mounting hole of the spray housing 105. The sealing strip 113 is made of silicone or fluororubber and has good airtightness and thermal insulation performance, ensuring effective isolation between the cooling area and the external environment and improving the overall energy efficiency ratio.

[0038] The heat dissipation cavity 114 is located outside the hot end of the semiconductor cooling chip 112, and is integrally formed with the spray shell 105 or tightly attached with thermally conductive silicone grease, for containing and guiding the dissipation of heat. The heat dissipation cavity 114 is provided with partitions 115 made of highly thermally conductive metal (such as aluminum or copper). The partitions 115 are fixed inside the cavity in an alternating or honeycomb structure, which not only enhances the structural strength, but also significantly increases the heat dissipation surface area and promotes the rapid diffusion of heat.

[0039] At least one circulating fan 116 is also installed inside the heat dissipation cavity 114. This fan is driven by an independent temperature control circuit and can automatically adjust its speed according to the operating temperature of the thermoelectric cooler 112. After the circulating fan 116 is started, it forces air to flow within the heat dissipation cavity 114, quickly carrying away the heat generated at the hot end and preventing the cooling efficiency from decreasing or the device from overheating and being damaged due to heat accumulation. In some embodiments, heat dissipation fins or a miniature liquid cooling circuit can also be added to the outside of the housing to further improve the heat dissipation capacity.

[0040] The control system intelligently adjusts the operating voltage and current of the semiconductor cooling chip 112 based on data collected by multimodal sensors (such as ambient temperature, target area temperature and humidity, spray flow rate, and set cooling targets) to dynamically control the cooling intensity. For example, during high-temperature periods in summer or in specific disinfection operations, the system can automatically activate the full-power cooling mode to output cold mist close to the dew point temperature; while in regular dust suppression operations, the cooling power can be turned off or reduced to save energy.

[0041] The fog cannon based on multimodal data control also includes a control module, which comprises a site photography unit 117, an area selection unit 118, a pedestrian density calculation unit 119, and a spray parameter generation unit 120. The site photography unit 117 is used to acquire site images; the area selection unit 118 is used to calculate the sunlit area based on the site images; the pedestrian density calculation unit 119 is used to calculate the pedestrian density based on the image of the sunlit area; and the spray parameter generation unit 120 is used to calculate the coordinates and range of the spray area based on the pedestrian density, and generate spray control parameters based on the coordinates and range.

[0042] The site imaging unit 117 consists of one or more high-definition wide-angle cameras (optionally equipped with infrared or thermal imaging sensors), mounted on the top of the fog cannon or on a gimbal structure above the spray housing 105. It features autofocus, image stabilization, and rain / fog penetration imaging capabilities. Its main function is to periodically or on-demand acquire real-time image information of the work site, covering an area of ​​up to several hundred square meters. Image data is transmitted via a high-speed interface to a local embedded processor or edge computing unit, providing raw visual input for subsequent intelligent analysis. In nighttime or low-light conditions, the system can automatically switch to infrared mode to ensure all-weather perception capabilities.

[0043] Based on the image data acquired by the site photography unit 117, the region selection unit 118 uses image processing algorithms (such as HSV color space analysis, light intensity gradient detection, shadow segmentation, etc.) to identify and divide the directly sunny and shaded areas in the current environment. This unit can combine timestamps and geographic coordinates, and call upon a built-in solar trajectory model for auxiliary judgment to improve recognition accuracy. Sunlit areas typically have higher surface temperatures and stronger thermal radiation effects, making them priority areas for misting and cooling. The system can mark the identified sunlit areas as polygonal or raster maps, serving as the basis for the next step of pedestrian density analysis. After determining the sunlit areas, the pedestrian density calculation unit 119 uses deep learning-based target detection algorithms (such as YOLO, SSD, or Mask R-CNN) to identify and locate human targets within the area. By real-time statistical analysis of parameters such as the number of pedestrians, distribution density, and movement trends in the image, the pedestrian density value per unit area (e.g., people / square meter) is calculated. This unit can also combine time series analysis to predict crowd gathering trends and identify peak periods or sudden gathering events, thereby enabling the early activation of spraying contingency plans. Furthermore, the system can be configured with privacy protection mechanisms to blur or anonymize identified facial information, complying with data security regulations.

[0044] The spray parameter generation unit 120 receives the output from the crowd density calculation unit 119 and, in conjunction with a preset spray strategy database (such as mist volume levels, spray angles, and wind speed settings corresponding to different crowd densities), automatically generates the optimal set of spray control parameters. Specifically, this includes: Determine the coordinates and coverage area of ​​the target area to be sprayed (based on geolocation and gimbal angle calculation). Set the opening and closing combination and atomization pressure of nozzle 106; Adjust the speed of fan 107 to match the required range and diffusion angle; Controlling the rotation angle of the first rotating seat 103 and the second rotating seat 104 ensures that the spray direction is precisely aimed at areas with high-density populations. Adjust the opening and closing degree of the adjustable nozzle 108 to control the mist flow pattern (concentrated jet or wide-angle diffusion). If equipped with a refrigeration system, the refrigeration function can be started and stopped in conjunction with the ambient temperature and the density of people.

[0045] All parameters are transmitted to each actuator via CAN bus or industrial Ethernet to achieve closed-loop control of "sensing-decision-execution".

[0046] The region selection unit 118 includes an image conversion subunit 122, a brightness threshold segmentation subunit 123, and a region generation subunit 124. The image conversion subunit 122 is used to convert an RGB color image into a grayscale image and eliminate image noise. The brightness threshold segmentation subunit 123 is used to calculate the global brightness mean and standard deviation based on the site image and set a dynamic threshold, and compare the pixel brightness value with the dynamic threshold to generate a mask. The region generation subunit 124 is used to eliminate small noise points, fill holes in the sunlight area, and retain continuous bright areas as sunlight-illuminated areas.

[0047] This sub-unit is responsible for converting the input RGB color image into a grayscale image and preprocessing the image to eliminate noise and improve the accuracy of subsequent segmentation.

[0048] Specific implementation steps: Convert the RGB three-channel image into a single-channel grayscale image using a weighted average method. gray (x, y), its calculation formula is as follows: I gray (x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y)Igray(x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y) Where R(x,y), G(x,y), and B(x,y) represent the red, green, and blue channel values ​​at pixel (x,y), respectively.

[0049] To reduce high-frequency noise in the image (such as sensor noise or illumination fluctuations), a Gaussian filter is used to smooth the grayscale image. I smooth (x,y)=I gray (x,y)∗G(x,y) Where * denotes the convolution operation, and G(x,y) is the Gaussian kernel function: Typically, σ=1 is chosen, and the filter kernel size is 5×5.

[0050] The brightness threshold segmentation subunit 123 dynamically calculates the segmentation threshold based on the brightness statistical features of the image to adapt to images under different lighting conditions and achieve adaptive binarization segmentation.

[0051] For the smoothed grayscale image I smooth Calculate its global mean μ and standard deviation σ I : Where M×N is the size of the image.

[0052] To enhance adaptability to different lighting environments, a dynamic threshold strategy combining the mean and standard deviation is adopted: T=μ+kσ I Where k is an empirical coefficient, which usually ranges from 0.5 to 1.5 and can be adjusted according to the actual scene (for example, a smaller value is used under strong light and a larger value is used under weak light).

[0053] The brightness value of each pixel is compared with the threshold T to generate a binary mask: Pixels with a value of 1 in the mask represent potential areas exposed to sunlight.

[0054] The region generation subunit 124 performs post-processing on the binary mask to remove noise interference, fill holes, and extract continuous bright areas as the final sunlight-illuminated areas.

[0055] Use morphological opening operations (erosion followed by dilation) to remove isolated noise points: M open =(M∘B)⊕B Where ∘ represents corrosion, ⊕ represents expansion, and B is the structural element (usually a 3×3 square core).

[0056] For M open Holes are filled in connected regions. Morphological closing operations or region filling algorithms are used: M filled =M open ⊕B∘B Alternatively, connected component analysis can be used to fill in the dark points within each bright area.

[0057] Retain the largest continuous bright area or filter regions by setting an area threshold, extract all continuous bright areas through connected component analysis, and filter according to area size: A i =Area(Di) Where Di is the i-th connected region, A i Given its pixel area. Only retain those satisfying A. i >A min Area (A) min The minimum effective area threshold (e.g., 50 pixels) is used to exclude small bright areas that are falsely detected.

[0058] After the above processing, a clean and continuous sunlight-illuminated area mask is obtained, which can be used for subsequent analysis or visualization.

[0059] The pedestrian density calculation unit 119 includes a pedestrian detection subunit 125, a pedestrian tracking subunit 126, a pedestrian density calculation subunit 127, and a density grading subunit 128. The pedestrian detection subunit 125 is used to perform pedestrian analysis in the sunlit area using YOLOv8 to obtain the bounding box and confidence score of each pedestrian. The pedestrian tracking subunit 126 is used to associate pedestrians in continuous frames based on the detection box and motion prediction, and output the unique ID and motion trajectory of each pedestrian. The pedestrian density calculation subunit 127 is used to calculate the pedestrian density by mapping the pixel area to the real world scale through camera calibration. The density grading subunit 128 is used to perform fog control grading based on the pedestrian density to obtain the density level.

[0060] The pedestrian detection subunit 125 is responsible for pedestrian identification within a given sunlit area. It employs a state-of-the-art YOLOv8 model to locate pedestrians in the image and evaluate their confidence scores. YOLOv8 is a deep learning-based object detection algorithm capable of processing video streams in real time and exhibiting high accuracy in pedestrian detection against complex backgrounds.

[0061] The specific operation involves inputting real-time video frames of the sunlit area. YOLOv8 is used to detect pedestrians in each frame, outputting the bounding boxes of the pedestrians and their corresponding confidence scores. A list of the location information and confidence scores of all pedestrians in each frame provides foundational data for subsequent tracking.

[0062] The pedestrian tracking subunit 126 tracks an individual's motion trajectory by associating the bounding boxes of pedestrians in consecutive frames. This not only helps determine each person's positional changes but also assigns a unique ID for long-term tracking.

[0063] The specific operation involves inputting bounding boxes and confidence scores from the pedestrian detection subunit 125. Kalman filters or algorithms such as SORT / DeepSORT are applied to combine bounding boxes and predicted motion to match pedestrians between different frames. This method effectively addresses tracking challenges in situations with brief occlusion and dense crowds. Each tracked pedestrian has a unique ID and their motion trajectory, including changes in location coordinates over time.

[0064] The pedestrian density calculation subunit 127 converts the pixel-level pedestrian distribution obtained from the video frame into density values ​​in the actual physical space. To achieve this, camera calibration must first be completed to establish a mapping relationship between pixel coordinates and real-world coordinates.

[0065] The specific operation involves inputting the pedestrian position coordinates provided by the pedestrian tracking subunit 126. Camera calibration is performed to obtain a transformation matrix. The transformation matrix is ​​then used to convert the pixel area to real-world area units. Pedestrian density is calculated based on the distribution of pedestrians within a specific area, typically expressed as people per square meter. The pedestrian density value for a specified time period is then selected.

[0066] Density grading subunit 128 categorizes pedestrian flow based on calculated density values ​​to support further decision-making, such as adjusting public safety measures or triggering alarms.

[0067] The specific operation involves inputting the density value provided by the crowd density calculation subunit 127. Multiple density levels (such as low, medium, and high) are defined, and corresponding thresholds are set. The calculated density value is then categorized into the appropriate level. The crowd density level in the current scenario can be used for automated control or other response mechanisms, such as automatically activating the atomizing cooling system to improve comfort when a "high" density level is reached.

[0068] The spray parameter generation unit 120 includes a fitting calculation subunit 129, an angle calculation subunit 130, and an atomization parameter calculation subunit 131. The fitting calculation subunit 129 is used to generate multiple circumscribed circles based on the sunlight irradiation area and the maximum spray range, and to obtain the center coordinate array and corresponding radius of the multiple circumscribed circles, wherein the corresponding radius is less than or equal to the maximum spray range. The angle calculation subunit 130 is used to output the pitch angle and azimuth angle of the rotating seat based on the transformation of the center coordinate array to the coordinate system of the sprayer rotating seat. The atomization parameter calculation subunit 131 is used to match the atomization particle size and flow rate based on the corresponding radius and density level.

[0069] The fitting calculation subunit 129 is based on the sunlight-irradiated area identified by the previous stage (represented as a binary mask or contour), combined with the maximum spray radius R of the fog cannon. max The target area is geometrically fitted to generate a set of circumcircles with optimal coverage, where each circumcircle represents a potential spray coverage unit.

[0070] The specific implementation steps are as follows: extract the boundary contours D of all connected regions from the mask image output by the region selection unit 118. i This can be achieved using edge detection algorithms (such as Canny) or OpenCV's findContours function.

[0071] For each contour D i The center O of the circumcircle is calculated using a minimum area circumcircle algorithm (such as the minimum enclosing circle algorithm). i =(x i ,y i ) and radius ri : (O i ,r i =MinEnclosingDircle(D i ) Requirement r i ≤R max If r i >R max Then, the large region is segmented (e.g., based on clustering or grid partitioning), and multiple sub-circles are fitted to each segment. This yields a set of spray target points: P={(x1,y1,r1),(x2,y2,r2),…,(x n ,y n ,r n )} Where r k ≤R max Ensure that each spraying action is within the equipment's capabilities.

[0072] The angle calculation subunit 130 transforms the spray target point (center of the circumcircle) in the image plane from the image coordinate system to the sprayer mechanical coordinate system, and calculates the pitch angle \thetaθ and azimuth angle \phiϕ required for the sprayer rotating seat to accurately point to the center of the target area.

[0073] Using the camera intrinsic matrix K and inverse perspective transformation, the pixel (u,v) is projected onto a 3D ray direction in the camera coordinate system. If the target point is known to be located on the ground plane (Z=0), it can be directly mapped using the homography matrix H. Where (X) w ,Y w () represents the ground position in the world coordinate system.

[0074] The transformation from world coordinates to the local coordinate system of the fog cannon is as follows: Let the installation position of the fog cannon be (Xs, Ys, Zs), then the relative displacement vector of the target point is: d=(X w -Xs,Y w −Ys,−Zs) The azimuth angle for calculating the rotation angle is: ϕ=arctan2(Y w -Ys,X w −Xs) ϕ represents the horizontal rotation angle (relative to due north or the front of the equipment).

[0075] The pitch angle is: This indicates the vertical elevation angle; a positive value indicates an upward tilt.

[0076] Output the (ϕ) corresponding to each group of target points i ,θ i This is used by the servo motor control system.

[0077] Atomization parameter calculation subunit 131 calculates the spray coverage requirements (circumscribed circle radius r) based on the spray coverage requirements. i Based on the on-site pedestrian density level L, the system dynamically matches the optimal atomization particle size (droplet diameter D) and spray flow rate Q to achieve a balance between comfort and energy efficiency.

[0078] Specific implementation steps: Spray particle size affects evaporation efficiency and perceived comfort. Smaller particle sizes (20–50 μm) evaporate quickly but are easily dispersed by wind; larger particle sizes (80–100 μm) provide strong cooling but result in a higher perceived humidity. The spray pattern adapts to the coverage radius. D=D min +(r i / R max (D) max -D min ) For example: D min =30μm, D max =90μm Simultaneously, adjustments are made based on density levels L∈{1,2,3} (low, medium, high): D=Dα L Where α1=1.2 (slightly larger in sparse areas), α2=1.0, and α3=0.8 (rapid cooling with fine mist in dense areas).

[0079] The flow rate is related to the coverage area and the cooling intensity. Let the recommended flow rate per unit area be q0 (e.g., 0.5 L / min / m²). 2 ),but: Q=q0πr i 2 βL βL is the density gain coefficient: β1=0.6 (low density, energy-saving mode), β2=1.0, β3=1.5 (high density, strong cooling).

[0080] Generate control vectors for each set of spray actions: F i =(ϕ i ,θ i D i Q i ) The signal is sent to the mist sprayer control system to achieve precise directional spraying.

[0081] The control module also includes a correction unit 121, which is used to acquire wind data of the irradiated area and correct the spray control parameters based on wind speed and wind direction.

[0082] The correction unit 121 includes a wind data acquisition subunit 132, a spray offset compensation subunit 133, and a mist particle size adjustment subunit 134. The wind data acquisition subunit 132 is used to acquire wind speed and wind direction angle in real time through wind speed and wind direction sensors. The spray offset compensation is used to calculate the spray angle offset based on the wind speed and wind direction angle, and adjust the pointing angle of the rotating seat. The mist particle size adjustment subunit 134 is used to match the corresponding atomization particle size based on the wind speed level to reduce dispersion loss.

[0083] Deploy high-precision ultrasonic wind speed and direction sensors or mechanical anemometers near the fog cannon, with a sampling frequency of ≥1Hz recommended.

[0084] Wind speed value v w The unit is m / s, which represents the speed of airflow. Wind direction angle ψ: Measured clockwise from true north (0°), in degrees (°). For example: ψ=0∘: North wind (blowing from north to south) ψ=90∘: East wind (blowing from east to west) Due to wind, the spray cloud will deviate from its trajectory during flight, causing the actual landing point to deviate from the target area. The spray deviation compensation subunit 133 establishes a wind-induced deviation model, calculates the compensation angle of the spray direction, and adjusts the azimuth angle of the spray nozzle's rotating seat to achieve headwind pre-deflection, ensuring that the spray covers the target area.

[0085] Based on empirical formulas or CFD simulation data fitting, the spray at its maximum range R... max The lateral drift distance at that point is approximately: d d =k d v w t f Where k d t is the drift coefficient; f For flight time, it can be estimated as tf≈R max / v0, where v0 is the initial velocity at the nozzle exit (e.g., 15–25 m / s).

[0086] d d =αv w (α≈0.5∼1.0s, empirical parameter) Let the wind direction angle be ψ and the azimuth angle of the target spray direction be ϕ. Then, the direction of the wind-induced offset is perpendicular to the wind direction. To counteract this offset, the spray direction needs to be adjusted against the wind by an angle, which simplifies to: Δϕ=βv w cos(ψ−ϕ) Where β is the compensation gain coefficient (e.g., β=0.8∘s / m), and ψ−ϕ is the angle between the wind direction and the jet direction.

[0087] The corrected azimuth angle is: ϕ new =ϕ+Δϕ If |Δϕ|>Δϕ max If the angle is 30°, an alarm will be triggered or a multi-point spraying strategy will be enabled.

[0088] This compensation mechanism enables an intelligent response that "the nozzle 106 deflects in the opposite direction of the wind," significantly improving coverage accuracy.

[0089] In strong winds, fine mist droplets are easily dispersed, evaporate too quickly, or become ineffectively dispersed, leading to water waste and reduced cooling efficiency. The mist particle size adjustment subunit 134 dynamically adjusts the atomization particle size according to the wind speed level to reduce wind-induced losses.

[0090] Wind speed is divided into multiple levels to match different atomization strategies: Wind speed level v_wvw (m / s) describe L0 (No wind) <2 Calm and stable, suitable for fine fog L1 (Gentle Breeze) 2–4 Mild impact L2 (stroke) 4–6 Noticeably drifting L3 (Strong Wind) >6 Strong interference requires coarse fog. Dynamically adjust droplet diameter D: The original droplet diameter D0 comes from the previous stage "atomization parameter calculation subunit 131", and is now corrected according to the wind speed: D final =D0γ(v w ) Where γ(v) w The wind speed gain factor can be set according to a range. That is, the stronger the wind, the coarser the fog droplets, thus enhancing wind resistance.

[0091] Since the increased particle size leads to a decrease in the number of droplets per unit volume, the flow rate should be appropriately increased to maintain the cooling effect. Q new =Q(D final / D0) 2 The correction unit 121 significantly improves the robustness and adaptability of the outdoor intelligent spray system under complex weather conditions by introducing wind sensing and dynamic compensation mechanisms. Second Embodiment This invention also provides a fogging method based on multimodal data control, comprising: S201 collected site photos; High-definition visible light cameras deployed at high or key locations on the site continuously collect real-time images. These cameras have a wide field of view, covering the main activity areas of the target region. The acquired images are color digital images, containing rich lighting and scene information, serving as the foundational data source for subsequent environmental status perception. The system can acquire images according to set time intervals or event-triggered mechanisms, ensuring the timeliness and stability of data updates.

[0092] S202 calculates the sunlit area based on site images; Image processing techniques were used to analyze the acquired site images and identify areas directly exposed to sunlight. This process first converted the color images to grayscale and then performed noise suppression and illumination equalization to improve image quality. Subsequently, an adaptive thresholding algorithm was employed to distinguish brighter areas in the image that were significantly brighter than their surroundings; these areas typically correspond to light spots formed by direct sunlight on the ground.

[0093] Further morphological operations were used to remove minor noise, fill in broken areas, and extract the outlines of continuous bright areas, ultimately generating the boundary ranges of one or more sun-exposed areas. These areas were considered target areas requiring focused cooling treatment, providing a spatial basis for subsequent spray control.

[0094] S203 calculates pedestrian density based on images of sunlit areas; Within the identified sunlit areas, further pedestrian detection and statistical analysis are performed to assess the level of crowd gathering in the area. The system employs an advanced deep learning object detection model to identify human targets in the images and obtain the location information of each pedestrian. Through continuous analysis of multiple frames of images, the system can also track pedestrian trajectories and associate their identities, avoiding duplicate counting.

[0095] Based on the number of pedestrians appearing in the sunlit area per unit time and their distribution range, the system calculates the pedestrian density of the area, i.e., the number of people per unit area. According to the preset density level classification standard, the area is divided into low-density, medium-density, and high-density levels for the formulation of differentiated spraying strategies.

[0096] S204 calculates the coordinates and range of the spray area based on the pedestrian density, and generates spray control parameters based on the coordinates and range.

[0097] By combining the spatial location of the sunlit area with the corresponding pedestrian density level, the system determines the actual spraying area that needs to be activated. For sunlit areas with high pedestrian density, the system sets them as the primary spraying target; while for unoccupied or low-density areas, the spray intensity is reduced or the spray is temporarily turned off, achieving on-demand response.

[0098] Based on the geometric characteristics of the spray area, such as its center position, coverage area, and shape, the system calculates the target coordinates and coverage radius that the sprayer should point towards. This coordinate information is then transformed and mapped to the physical installation coordinate system of the spraying equipment, generating spatial pointing instructions that the equipment can recognize.

[0099] Based on this, the system further generates complete spray control parameters, including spray direction (azimuth and pitch angles), spray intensity, droplet size, and spray duration. The setting of control parameters comprehensively considers the area size, population density, and environmental comfort requirements, ensuring that the spray can effectively cool the area without causing excessive humidification or wasting resources.

[0100] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A fog cannon based on multimodal data control, comprising a support base and movable wheels, wherein the movable wheels are disposed at the bottom of the support base, characterized in that, It also includes a first rotating seat, a second rotating seat, a spray housing, multiple nozzles, a fan, and an adjustable nozzle. The first rotating seat is rotatably mounted on the support base, and the second rotating seat is rotatably mounted on the first rotating seat. The spray housing is fixed on the second rotating seat. The multiple nozzles are disposed inside the spray housing. The fan is disposed on one side of the nozzles. The adjustable nozzle includes a sliding conical sleeve, a control component, and a limiting ring. The sliding conical sleeve is slidably disposed on one side of the spray housing. The control component is connected to the sliding conical sleeve. The limiting ring is fixed on the sliding conical sleeve and matches the sealing ring on the spray housing.

2. The fogging device based on multimodal data control as described in claim 1, characterized in that, The fogging device based on multimodal data control also includes a cooling structure, which is disposed inside the spray housing.

3. The fogging device based on multimodal data control as described in claim 2, characterized in that, The cooling structure includes multiple semiconductor cooling chips, a sealing strip, a heat dissipation cavity, a partition, and a circulating fan. The multiple semiconductor cooling chips are disposed on one side of the nozzle for cooling the spray. The sealing strip is disposed between the semiconductor cooling chips and the spray housing. The heat dissipation cavity is disposed on the outside of the semiconductor cooling chips. The partition is fixed inside the heat dissipation cavity. The circulating fan is disposed inside the heat dissipation cavity.

4. The fogging device based on multimodal data control as described in claim 3, characterized in that, The fog cannon based on multimodal data control also includes a control module, which includes a site photography unit, an area selection unit, a crowd density calculation unit, and a spray parameter generation unit. The site photography unit is used to capture site images; The area selection unit is used to calculate the sunlit area based on the site image; The crowd density calculation unit is used to calculate crowd density based on images of sunlit areas; The spray parameter generation unit is used to calculate the coordinates and range of the spray area based on the pedestrian density, and to generate spray control parameters based on the coordinates and range.

5. The fogging device based on multimodal data control as described in claim 4, characterized in that, The region selection unit includes an image conversion subunit, a brightness threshold segmentation subunit, and a region generation subunit; The image conversion subunit is used to convert RGB color images into grayscale images and eliminate image noise; The brightness threshold segmentation subunit is used to calculate the global brightness mean and standard deviation based on the site image and set a dynamic threshold, and compare the pixel brightness value with the dynamic threshold to generate a mask. The region generation sub-unit is used to eliminate small noise points and fill the voids in the sunlight area, while retaining continuous bright areas as the sunlight-illuminated area.

6. The fogging device based on multimodal data control as described in claim 5, characterized in that, The pedestrian density calculation unit includes a pedestrian detection subunit, a pedestrian tracking subunit, a pedestrian density calculation subunit, and a density grading subunit; The pedestrian detection subunit is used to perform pedestrian flow analysis in the sunlit area using YOLOv8 to obtain the bounding box and confidence score of each pedestrian. The pedestrian tracking subunit is used to output a unique ID and motion trajectory for each pedestrian based on the detection box and motion prediction associated with pedestrians in continuous frames. The crowd density calculation subunit is used to calculate crowd density by mapping pixel area to the real-world scale through camera calibration. The density grading subunit is used to perform fog control grading based on pedestrian flow density to obtain density levels.

7. The fogging device based on multimodal data control as described in claim 6, characterized in that, The spray parameter generation unit includes a fitting calculation subunit, an angle calculation subunit, and an atomization parameter calculation subunit; The fitting calculation subunit is used to generate multiple circumcircles based on the sunlight-irradiated area and the maximum spray range, and to obtain the center coordinate array and corresponding radius of the multiple circumcircles, wherein the corresponding radius is less than or equal to the maximum spray range; The angle calculation subunit is used to output the pitch and azimuth angles of the rotating seat based on the transformation of the center coordinate array to the coordinate system of the spray nozzle rotating seat; The atomization parameter calculation subunit is used to match the atomization particle size and flow rate based on the corresponding radius and density level.

8. The fogging device based on multimodal data control as described in claim 7, characterized in that, The control module also includes a correction unit, which is used to acquire wind data of the irradiated area and correct the spray control parameters based on wind speed and wind direction.

9. The fogging device based on multimodal data control as described in claim 8, characterized in that, The correction unit includes a wind data acquisition subunit, a spray offset compensation subunit, and a mist particle size adjustment subunit; The wind data acquisition subunit is used to acquire wind speed and wind direction angle in real time through wind speed and wind direction sensors; The spray offset compensation is used to calculate the spray angle offset based on the wind speed value and wind direction angle, and to adjust the pointing angle of the rotating seat. The mist particle size adjustment subunit is used to match the corresponding atomization particle size based on the wind speed level in order to reduce drift loss.

10. A fogging method based on multimodal data control, employing a fogging device based on multimodal data control as described in claim 9, characterized in that, include: Collect site photos; Calculate the sunlit area based on site images; Calculate pedestrian density based on images of sunlit areas; The coordinates and range of the spray area are calculated based on the pedestrian density, and spray control parameters are generated based on the coordinates and range.

Images