Three-dimensional raindrop trajectory identification system and identification method suitable for wind-rain coupling scenarios

Abstract

The present application relates to the technical field of water and soil conservation, and particularly relates to a three-dimensional raindrop trajectory identification system and method suitable for a wind-rain coupling scenario, the three-dimensional raindrop trajectory identification system comprising a wind field information acquisition device, a multi-view imaging device, an image processing module, a particle parameter calculation module and a central control module. Wind speed and direction data and multi-view raindrop image sequences are acquired in real time, and the wind field data is integrated into the image processing algorithm as a physical constraint to assist in matching and tracking the raindrop target, and an accurate three-dimensional motion trajectory is constructed. The particle parameter calculation module combines the trajectory and wind field data to analyze physical parameters such as raindrop particle size, speed, incident angle and kinetic energy. The present application breaks through the limitation that traditional observation technology cannot effectively perceive the wind-rain coupling effect, significantly improves the accuracy and robustness of target identification and trajectory reconstruction in complex environments, and provides quantitative data support for wind-driven rain mechanism research, water and soil conservation and engineering protection.

Description

Technical Field

[0001] This invention relates to the field of soil and water conservation technology, and in particular to a three-dimensional raindrop trajectory recognition system and method suitable for wind and rain coupled scenarios. Background Technology

[0002] During natural rainfall, the interference effect of wind is particularly significant, especially under strong wind conditions. Raindrops undergo noticeable motion shifts and morphological changes during their fall, altering their velocity distribution and spatial trajectory, creating a "wind-driven rain" phenomenon. This wind-rain coupling effect not only changes the distribution of raindrop impact points and incident angles but also significantly impacts rainfall kinetic energy, splash erosion capacity, and surface erosion intensity, thus profoundly affecting soil erosion processes and the migration and diffusion mechanisms of non-point source pollutants. Therefore, accurately capturing the three-dimensional motion trajectory of raindrops under wind-driven rain conditions has significant scientific research and engineering application value for revealing erosion mechanisms, eco-hydrological effects, and optimizing the layout of soil and water conservation projects under wind-rain coupling scenarios.

[0003] However, existing raindrop trajectory recognition technologies still have significant limitations. Traditional rainfall monitoring equipment (such as tipping bucket rain gauges) is based on the ideal assumption of vertical rainfall and cannot reflect the actual movement path and angular changes of raindrops under the influence of wind. Most observation methods based on high-speed photography or image recognition are still limited to a two-dimensional perspective and are easily affected by factors such as field of view limitation, background interference, and raindrop overlap, resulting in image blurring, segmentation difficulties, and trajectory breaks, which seriously restricts the accuracy of raindrop recognition and parameter extraction. Although methods such as laser light curtains and structured background boards can partially improve imaging quality, they are still insufficient in multi-target tracking, three-dimensional continuous reconstruction, and adaptability to complex environments, making it difficult to achieve high-precision restoration of the complete movement trajectory of raindrops under combined wind and rain conditions.

[0004] In recent years, with the rapid development of high-speed imaging technology, stereo vision, and artificial intelligence methods, research has begun to explore the reconstruction of 3D raindrop trajectories based on multi-view vision, structured light projection, and other multi-angle imaging methods. However, these methods are mostly focused on applications under windless or indoor static conditions, and have not yet achieved synchronous perception and fusion analysis of wind field data, lacking the ability to jointly analyze the wind-rain coupling mechanism. Therefore, there is an urgent need to construct an integrated system and method that can synchronously perceive wind field information in real wind and rain environments, achieve accurate reconstruction of 3D raindrop trajectories, and invert physical parameters, in order to support high-precision observation and mechanistic analysis of wind-driven rain processes. Summary of the Invention

[0005] This invention provides a three-dimensional raindrop trajectory recognition system and method suitable for wind and rain coupling scenarios, which solves the shortcomings of existing technologies in accurately recognizing three-dimensional raindrop trajectories and particle parameter motion characteristics under wind action, and realizes high-precision recognition and dynamic three-dimensional reconstruction of raindrop trajectories under multi-source data fusion, providing reliable technical means and data support for wind-driven rain research.

[0006] This invention provides a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios, including a wind field information acquisition device, a multi-view imaging device, an image processing module, a particle parameter calculation module, and a central control module. The wind field information acquisition device is set in the target area to collect and output wind speed and direction data of the target area in real time. The multi-view imaging device is set in the target area to collect and output raindrop image sequences of the target area from multiple perspectives in real time. The image processing module is connected to the wind field information acquisition device and the multi-view imaging device to process the raindrop images collected by the multi-view imaging device. The system uses a sequence of images and, based on the wind speed and direction data output by the wind field information acquisition device, identifies and matches raindrop targets to construct a three-dimensional raindrop trajectory. The particle parameter calculation module connects the wind field information acquisition device and the image processing module, receiving the three-dimensional raindrop trajectory data constructed by the image processing module and calculating and extracting the physical characteristic parameters of the raindrops based on the wind speed and direction data output by the wind field information acquisition device. The central control module connects to the wind field information acquisition device, the multi-view imaging device, the image processing module, and the particle parameter calculation module for synchronous control and communication management.

[0007] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios further includes a laser array auxiliary module disposed in the target area. The laser array auxiliary module is connected to the image processing module. The laser array auxiliary module is used to acquire the spatial coordinate information of raindrops in the target area to calibrate the three-dimensional raindrop trajectory constructed by the image processing module.

[0008] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The laser array auxiliary module includes a line laser generator, a deflectable scanning galvanometer, and a synchronization and control unit. The line laser generator is used to generate a static laser plane. The scanning galvanometer is arranged opposite to the line laser generator, and the scanning galvanometer deflects the static laser plane generated by the line laser generator to form a layered spatial laser plane in the target area. The synchronization and control unit is connected to the central control module and the scanning galvanometer, respectively. The synchronization and control unit is used to receive the synchronization signal from the central control module and control the deflection angle and deflection speed of the scanning galvanometer.

[0009] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The system further includes a data processing and visualization module, which comprises a data fusion submodule, a physical parameter calculation engine, a visualization rendering engine, and a user interaction and output interface submodule. The data fusion submodule is connected to the wind field information acquisition device, the image processing module, and the particle parameter calculation module, respectively, to receive and fuse wind speed and direction data, raindrop physical characteristic parameters, and three-dimensional raindrop trajectories of the target area. The physical parameter calculation engine is connected to the data fusion submodule and is used to calculate and generate three-dimensional raindrop trajectory dynamics, wind field superposition data, and the time / space variation trend of raindrop physical characteristic parameters. The visualization rendering engine is connected to the physical parameter calculation engine and is used to generate three-dimensional raindrop trajectory dynamics, wind field superposition maps, and time / space variation trend maps of the physical characteristic parameters of each raindrop. The user interaction and output interface submodule includes a human-computer interaction interface and an output interface. The human-computer interaction interface is connected to the visualization rendering engine for visual graphic output; the output interface is connected to the physical parameter calculation engine for data export.

[0010] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The wind field information acquisition device includes a three-dimensional ultrasonic anemometer or a wind vector sensor. The three-dimensional ultrasonic anemometer or the wind vector sensor is used to acquire and output the wind speed and wind direction of the target area.

[0011] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The multi-view imaging device includes at least two visual imaging devices and a background plate. The optical axes of the two visual imaging devices are intersected and intersect at the center of the target area. The background plate is located on the opposite side of the two visual imaging devices and serves as the imaging background of the two visual imaging devices. The background plate is provided with calibration scale.

[0012] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The multi-view imaging device further includes an LED lighting device located laterally to the target area, the LED lighting device being used to illuminate raindrops entering the target area from the side; the visual imaging device is a high-speed camera with a frame rate of not less than 5000fps.

[0013] According to the present invention, a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios is provided. The image processing module includes an image preprocessing submodule, a target detection and segmentation submodule, a multi-target tracking submodule, and a stereo matching and three-dimensional reconstruction submodule. The image preprocessing submodule is connected to the multi-view imaging device and is used to acquire multiple frames of raindrop images collected by the multi-view imaging device, and to perform denoising, enhancement, and background extraction on the multiple frames of raindrop images. The target detection and segmentation submodule is connected to the image preprocessing submodule and is used to identify raindrop targets in each preprocessed frame of raindrop image. The system locates the spatial coordinates and time sequence of the raindrops. The multi-target tracking submodule is connected to the target detection and segmentation submodule and is used to associate the same raindrop across frames in multiple raindrop images to form a two-dimensional trajectory segment of the raindrop. The stereo matching and three-dimensional reconstruction submodule is connected to the wind field information acquisition device and the multi-target tracking submodule. Based on the multi-view information of the multi-view imaging device and the wind speed and wind direction data output by the wind field information acquisition device, the stereo matching and three-dimensional reconstruction submodule fuses the formed two-dimensional trajectory segments of the raindrops, calculates the three-dimensional spatial coordinates of the raindrops, and constructs a three-dimensional raindrop trajectory.

[0014] This invention also provides a method for recognizing three-dimensional raindrop trajectories in wind-rain coupling scenarios, comprising:

[0015] The wind speed and direction data of the target area are collected and output in real time through the wind field information acquisition device.

[0016] The system uses a multi-view imaging device to acquire and output a sequence of raindrop images of the target area in real time from multiple perspectives.

[0017] The spatial position and time signal of raindrops crossing the laser plane are recorded in real time by a laser array auxiliary module.

[0018] The image processing module processes the raindrop image sequence of the target area and the spatial and temporal signals of the raindrops crossing the laser plane to identify and match raindrop targets and construct a three-dimensional raindrop trajectory.

[0019] By combining the wind speed and direction data of the target area output by the particle parameter calculation module with the wind field information acquisition device, the constructed three-dimensional raindrop trajectory is dynamically analyzed, and the physical characteristic parameters of the raindrop are calculated and extracted.

[0020] The data processing and visualization modules are integrated for calculation, and the dynamic three-dimensional raindrop trajectory, wind field superimposed data, and the changing trend of raindrop physical property parameters over time and space are visualized.

[0021] According to the present invention, a three-dimensional raindrop trajectory recognition method suitable for wind and rain coupled scenarios is provided. The method involves processing the raindrop image sequence of the target area and the spatial position and temporal signal of the raindrop crossing the laser plane based on the image processing module, identifying and matching the raindrop target, and constructing the three-dimensional raindrop trajectory, including:

[0022] The image preprocessing submodule acquires multiple frames of raindrop images collected by the multi-view imaging device, and performs noise reduction, enhancement, and background extraction on the raindrop images.

[0023] The target detection and segmentation submodule is used to identify raindrop targets in each preprocessed raindrop image and locate their spatial coordinates and time series.

[0024] The multi-target tracking submodule associates the same raindrop across multiple raindrop images to form a two-dimensional trajectory segment of the raindrop.

[0025] The stereo matching and 3D reconstruction submodule uses multi-view information from a multi-view imaging device and wind speed and direction data from a wind field information acquisition device. It combines the spatial position and time signal of raindrops crossing the laser plane to fuse the formed two-dimensional trajectory segments of raindrops, calculate the three-dimensional spatial coordinates of the raindrops, and construct the three-dimensional raindrop trajectory.

[0026] This invention provides a 3D raindrop trajectory recognition system suitable for wind-rain coupling scenarios. By integrating real-time wind field data with multi-view visual imaging, it overcomes the limitations of traditional rain gauges or 2D photography techniques in sensing raindrop motion deviation under wind force, accurately reconstructing the true 3D trajectory and spatial distribution of raindrops in wind-rain coupling effects. By introducing wind field as prior physical knowledge, it greatly enhances the robustness and accuracy of multi-target matching and 3D reconstruction algorithms in complex wind-rain environments, effectively solving the recognition difficulties and tracking loss problems caused by raindrop overlap, motion blur, and trajectory intersection, and overcoming the shortcomings of pure visual methods in adaptability. This 3D raindrop trajectory recognition system not only outputs high-precision trajectory images but also provides quantitative particle dynamics parameters, providing reliable data support for research on wind-driven rain erosion mechanisms, optimization of soil and water conservation measures, and assessment of building wind and rain loads, achieving a significant leap from traditional 2D qualitative observation to 3D quantitative analysis. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0028] Figure 1This is a schematic diagram of the principle of the three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios provided by the present invention.

[0029] Figure 2 This is a flowchart illustrating the three-dimensional raindrop trajectory recognition method for wind and rain coupled scenarios provided by the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0031] The following description, in conjunction with the accompanying drawings, describes a three-dimensional raindrop trajectory recognition system and method applicable to wind and rain coupled scenarios.

[0032] One embodiment of the present invention provides a three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios, see [link to relevant documentation]. Figure 1 As shown, a 3D raindrop trajectory recognition system suitable for wind and rain coupled scenarios includes a wind field information acquisition device, a multi-view imaging device, an image processing module, a particle parameter calculation module, and a central control module. The wind field information acquisition device is set in the target area to collect and output wind speed and direction data of the target area in real time. The multi-view imaging device is set in the target area to collect and output raindrop image sequences of the target area from multiple perspectives in real time. The image processing module is connected to the wind field information acquisition device and the multi-view imaging device to process the raindrop image sequences collected by the multi-view imaging device, and to identify and match raindrop targets and construct 3D raindrop trajectories based on the wind speed and direction data output by the wind field information acquisition device. The particle parameter calculation module is connected to the wind field information acquisition device and the image processing module to receive the 3D raindrop trajectory data constructed by the image processing module, and to calculate and extract the physical characteristic parameters of the raindrops based on the wind speed and direction data output by the wind field information acquisition device. The central control module is connected to the wind field information acquisition device, the multi-view imaging device, the image processing module, and the particle parameter calculation module for synchronous control and communication management.

[0033] This embodiment of a 3D raindrop trajectory recognition system suitable for wind and rain coupled scenarios comprises five core modules. The first module, a wind field information acquisition device deployed in the target area (such as a 3D ultrasonic anemometer or wind vector sensor to acquire and output wind speed and direction data), is responsible for real-time capture and output of high spatiotemporal resolution wind speed and direction data. The second module, a multi-view imaging device typically composed of two or more high-speed cameras, synchronously acquires high-speed image sequences of falling raindrops from different angles to ensure image quality. The third module, the data processing core of the system, first preprocesses and segments the acquired image sequences, then introduces real-time wind field data as a physical constraint to assist the algorithm in matching and tracking raindrop targets across different viewpoints. Finally, it reconstructs the trajectory of the raindrops in 3D space using stereoscopic vision principles. The fourth module, a particle parameter calculation module, further analyzes and calculates key physical parameters such as raindrop size, velocity, incident angle, and kinetic energy based on the 3D trajectory and synchronized wind field data. The entire system is uniformly scheduled by a central control module to ensure millisecond-level time synchronization and efficient data flow coordination among the hardware devices during the acquisition process.

[0034] It is understandable that this 3D raindrop trajectory recognition system, applicable to wind-rain coupling scenarios, overcomes the limitations of traditional rain gauges or 2D photography techniques in sensing raindrop movement deviations under wind force by integrating real-time wind field data and multi-view visual imaging. It can accurately reconstruct the true 3D trajectory and spatial distribution of raindrops in the wind-rain coupling effect. By introducing wind field as prior physical knowledge, the robustness and accuracy of multi-target matching and 3D reconstruction algorithms in complex wind-rain environments are greatly enhanced. This effectively solves the recognition difficulties and tracking loss problems caused by raindrop overlap, motion blur, and trajectory intersections, overcoming the shortcomings of purely visual methods in terms of adaptability. This 3D raindrop trajectory recognition system not only outputs high-precision trajectory images but also provides quantitative particle dynamics parameters, providing reliable data support for research on wind-driven rain erosion mechanisms, optimization of soil and water conservation measures, and assessment of building wind and rain loads. It represents a significant leap from traditional 2D qualitative observation to 3D quantitative analysis.

[0035] In some embodiments of the three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios of the present invention, the three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios further includes a laser array auxiliary module disposed in the target area. The laser array auxiliary module is connected to the image processing module and is used to obtain the spatial coordinate information of the raindrop in the target area to calibrate the three-dimensional raindrop trajectory constructed by the image processing module.

[0036] It is understood that, in this embodiment, the 3D raindrop trajectory recognition system for wind and rain coupled scenarios of the present invention further integrates a laser array auxiliary module. When raindrops fall and pass through the layered laser plane generated by the laser array auxiliary module, strong Mie scattering occurs, resulting in a bright spot in the image sequence acquired by the high-speed camera, with a brightness much higher than the environmental background. The spatial position of these spots corresponds to the known and precisely calibrated coordinates of the laser plane. When the image processing module executes the 3D reconstruction algorithm, it first identifies and extracts these laser spots. Subsequently, it uses these known absolute spatial coordinates as a high-precision reference benchmark to perform spatial correction and calibration of the 3D raindrop trajectory calculated by binocular parallax. This embodiment effectively compensates for the calculation deviation caused by camera calibration error, lens distortion, or environmental interference by integrating the laser array auxiliary module, ultimately significantly improving the absolute accuracy of the raindrop 3D spatial position data and the overall robustness of the system.

[0037] In some specific examples, the laser array auxiliary module includes a line laser generator, a deflectable scanning galvanometer, and a synchronization and control unit. The line laser generator is used to generate a static laser plane. The scanning galvanometer is positioned opposite to the line laser generator and deflects the static laser plane generated by the line laser generator to form a layered spatial laser plane in the target area. The synchronization and control unit is connected to the central control module and the scanning galvanometer, respectively. The synchronization and control unit is used to receive the synchronization signal from the central control module and control the deflection angle and deflection speed of the scanning galvanometer.

[0038] Understandably, in this specific example, the laser array auxiliary module employs a dynamic scanning structure, the core of which is composed of three precision components working together. A line laser generator serves as the light source, emitting light to form a highly collimated static laser plane; a deflectable scanning galvanometer is precisely positioned opposite the line laser generator, its lens capable of high-speed and precise deflection on two rotation axes; and the synchronization and control unit acts as the "brain" of this module, connected to the system's central control module via a communication interface, and directly driving and controlling the movement of the scanning galvanometer.

[0039] Specifically, during implementation, the synchronization and control unit receives synchronization trigger signals and scanning parameters from the central control module. Based on these instructions, it drives the mirrors of the scanning galvanometer to deflect periodically. The static laser plane generated by the line laser generator is incident on the continuously moving mirrors. Through optical reflection, a static laser line is rapidly scanned across the target observation area, forming a dynamic, layered laser curtain with a larger coverage area composed of continuous scan lines. This allows a single line laser source to achieve calibration coverage over a large space. The central control module ensures that each scanning position of the galvanometer strictly corresponds to the exposure time of the high-speed camera. This ensures that the precise three-dimensional spatial coordinates of any bright spot generated by a raindrop crossing the laser plane can be calculated and recorded in real time, providing crucial spatial reference points for subsequent image matching and 3D reconstruction.

[0040] In some embodiments of the three-dimensional raindrop trajectory recognition system for wind-rain coupled scenarios of the present invention, the system further includes a data processing and visualization module. This module comprises a data fusion submodule, a physical parameter calculation engine, a visualization rendering engine, and a user interaction and output interface submodule. The data fusion submodule is connected to the wind field information acquisition device, the image processing module, and the particle parameter calculation module, respectively, to receive and fuse wind speed and direction data of the target area, physical characteristic parameters of raindrops, and three-dimensional raindrop trajectories. The physical parameter calculation engine is connected to the data fusion submodule and is used to calculate and generate three-dimensional raindrop trajectory dynamics, wind field superposition data, and the time / space variation trend of raindrop physical characteristic parameters. The visualization rendering engine is connected to the physical parameter calculation engine and is used to generate three-dimensional raindrop trajectory dynamics, wind field superposition maps, and time / space variation trend maps of the physical characteristic parameters of each raindrop. The user interaction and output interface submodule includes a human-computer interaction interface and an output interface. The human-computer interaction interface is connected to the visualization rendering engine for visual graphic output; the output interface is connected to the physical parameter calculation engine for data export.

[0041] It is understood that, in this embodiment, the 3D raindrop trajectory recognition system for wind and rain coupled scenarios of the present invention further integrates a dedicated data processing and visualization module. This module, as the central hub for information processing and presentation, consists of four functionally defined sub-units. The data fusion sub-module, as the input hub, is connected to the wind field information acquisition device, image processing module, and particle parameter calculation module through data interfaces, respectively, and is responsible for receiving and spatiotemporally aligning multi-source heterogeneous data. The physical parameter calculation engine is connected to the data fusion sub-module and serves as the core calculation unit, incorporating various physical models and statistical algorithms. The visualization rendering engine is connected to the calculation engine and possesses powerful graphics processing and rendering capabilities. Finally, the user interaction and output interface sub-module provides an external channel for the entire module, connecting to the visualization rendering engine, providing a graphical user interface (GUI), and supporting multiple data export protocols.

[0042] During implementation, the data fusion submodule first receives, decodes, and timestamp-matches the synchronously acquired raw wind vector data, 3D raindrop trajectory coordinate set, and physical parameter sequence, fusing them into a unified multidimensional dataset. Subsequently, the physical parameter calculation engine performs secondary deep calculations on this fused data, such as dynamizing the trajectory data, overlaying and fusing wind field vectors with raindrop motion trajectories, and statistically analyzing the evolution trends of parameters like kinetic energy and particle size over time or spatial location. The visualization rendering engine then converts these calculation results into intuitive graphical representations, generating and rendering 3D raindrop motion animations, wind-rain coupling overlay effect diagrams, and various parameter change curves in real time. Finally, through the user interaction and output interface submodule, users can observe and manipulate these visualization results in real time on the user interface, and can choose to export the raw data, processing results, or analysis charts in standard formats (such as CSV, images, and videos) for further professional analysis or report writing.

[0043] In some embodiments of the three-dimensional raindrop trajectory recognition system applicable to wind and rain coupled scenarios of the present invention, the multi-view imaging device includes at least two visual imaging devices and a background plate. The optical axes of the two visual imaging devices are intersected and intersect at the center of the target area. The background plate is located on the opposite side of the two visual imaging devices and serves as the imaging background of the two visual imaging devices. The background plate is provided with calibration scale.

[0044] The multi-view imaging device in this embodiment employs a precisely optically designed stereo vision structure. Its core hardware includes at least two high-frame-rate visual imaging devices (such as high-speed cameras) and a background plate corresponding to the high-frame-rate visual imaging devices. The two cameras are arranged separately at a certain baseline distance, with their optical axes intersecting and precisely converging at the center point of the target observation area, forming a stable stereo vision measurement baseline. The background plate is placed on the opposite side of these cameras, facing the camera lenses, serving as a unified imaging background. The surface of the background plate is dark (such as dark blue or black), and clearly defined calibration scales (such as horizontal-vertical scale lines, grid scales, or fan-shaped scales) are drawn on it.

[0045] During implementation, the system first uses the calibration scale on the background plate to jointly calibrate the two cameras, accurately calculating the internal parameters of each camera (such as focal length and distortion coefficient) and their relative position and orientation (i.e., external parameters), thereby establishing a precise mathematical transformation relationship between image pixel coordinates and three-dimensional world coordinates. During data acquisition, raindrops fall within the target area, and the two cameras simultaneously acquire images from different perspectives. The dark background plate maximizes the contrast between the raindrops and the background, and the scale on the plate not only plays a role in the initial calibration but also serves as a reference for the image coordinate system in subsequent measurements. Then, the image processing module uses the pre-calibrated camera parameters to perform stereo matching on the two-dimensional image coordinates of the same raindrop captured by the two cameras, calculating the precise three-dimensional spatial coordinates of the raindrop within the intersection area of ​​the two cameras' fields of view using triangulation principles.

[0046] To further improve imaging quality, some embodiments of the multi-view imaging device also integrate a dedicated lighting system. This system consists of LED lighting devices arranged to the side of the target area. The light emission direction is precisely adjusted, and a strong directional collimated backlighting method is used to illuminate the raindrops entering the target area from the side to highlight the trajectory and shape of the raindrops. The LED lighting devices can evenly and fully illuminate the raindrops entering the observation area from the side using a strong directional collimated backlighting method.

[0047] Furthermore, the visual imaging equipment used in the device consists of high-speed cameras with a frame rate of no less than 5000 fps. These cameras possess microsecond-level exposure times and highly sensitive image sensors. During implementation, side-illumination devices emit high-intensity, highly directional cold light sources, illuminating raindrops from a direction perpendicular to the camera's optical axis. This side-illumination method, combined with a dark background, creates an ideal dark-field imaging effect: raindrops appear as bright, clearly defined highlights in the image due to the parallel light beams, while the background remains dark, greatly enhancing the contrast between the target and the background and solving the problem of blurred edges and difficulty in segmentation of raindrop images. Simultaneously, the high frame rate acquisition of no less than 5000 fps ensures that even high-speed moving raindrops can be continuously and clearly captured, effectively avoiding motion blur. The ultra-short exposure time effectively "freezes" the moment of raindrop movement, making each frame clear and sharp, providing crucial high-quality raw data for subsequent high-precision target detection, tracking, and 3D reconstruction based on image sequences.

[0048] In some embodiments of the three-dimensional raindrop trajectory recognition system applicable to wind and rain coupled scenarios of the present invention, the image processing module includes an image preprocessing submodule, a target detection and segmentation submodule, a multi-target tracking submodule, and a stereo matching and three-dimensional reconstruction submodule. The image preprocessing submodule is connected to a multi-view imaging device to acquire multiple frames of raindrop images collected by the multi-view imaging device, and to perform denoising, enhancement, and background extraction on the multiple frames of raindrop images. The target detection and segmentation submodule is connected to the image preprocessing submodule to identify raindrop targets in each preprocessed frame of raindrop image and locate their spatial coordinates and time sequence. The multi-target tracking submodule is connected to the target detection and segmentation submodule to associate the same raindrop across frames in multiple raindrop images to form two-dimensional trajectory segments of raindrops. The stereo matching and three-dimensional reconstruction submodule is connected to a wind field information acquisition device and a multi-target tracking submodule. Based on the multi-view information of the multi-view imaging device and the wind speed and direction data output by the wind field information acquisition device, the stereo matching and three-dimensional reconstruction submodule fuses the formed two-dimensional trajectory segments of raindrops, calculates the three-dimensional spatial coordinates of the raindrops, and constructs a three-dimensional raindrop trajectory.

[0049] It is understood that the image processing module in this embodiment adopts a hierarchical, pipelined processing architecture, consisting of four functionally specialized and sequentially connected sub-modules. The multi-view imaging device, as the data input end, is directly connected to the image preprocessing sub-module to receive raw image data; the target detection and segmentation sub-module is connected to the preprocessing sub-module, focusing on target recognition in a single frame image; the multi-target tracking sub-module is connected to the target detection and segmentation sub-module, responsible for cross-frame correlation; finally, the stereo matching and 3D reconstruction sub-module, as the output end, not only receives 2D trajectory data from the multi-target tracking sub-module but also receives real-time data streams from the wind field information acquisition device.

[0050] In the implementation process, the image preprocessing submodule first performs denoising, contrast enhancement, and background subtraction on the input high-speed image sequence to maximize the highlighting of raindrop targets. Subsequently, the target detection and segmentation submodule (typically based on deep learning models such as YOLO or image segmentation algorithms) analyzes each processed frame, accurately identifying the contour of each raindrop target and outputting its pixel coordinates and timestamp. The multi-target tracking submodule (using algorithms such as SORT or DeepSORT) associates the same raindrop across consecutive frames, forming a two-dimensional motion trajectory segment for each raindrop from a single camera viewpoint. Finally, the stereo matching and 3D reconstruction submodule performs the core operation: receiving matched two-dimensional trajectory segments from different camera viewpoints and simultaneously introducing real-time wind speed and direction data as key physical constraints. These constraints greatly assist the algorithm in making correct matching decisions between different viewpoints and serve as optimization targets, improving the physical rationality of the reconstructed trajectory. Finally, by using the principle of binocular parallax triangulation, the two-dimensional coordinates are fused and calculated to accurately reconstruct the continuous position sequence of raindrops in three-dimensional space, and a high-fidelity three-dimensional motion trajectory that is subjected to wind field disturbance is constructed.

[0051] In another aspect, the present invention provides a three-dimensional raindrop trajectory recognition method suitable for wind-rain coupled scenarios. This method can be implemented based on any of the above embodiments or examples of a three-dimensional raindrop trajectory recognition system suitable for wind-rain coupled scenarios. In some embodiments of the three-dimensional raindrop trajectory recognition method suitable for wind-rain coupled scenarios of the present invention, see [link to relevant documentation]. Figure 2 As shown, the three-dimensional raindrop trajectory recognition method applicable to wind and rain coupled scenarios includes the following steps S1 to S4.

[0052] S1. Real-time acquisition and output of wind speed and direction data for the target area via a wind field information acquisition device. Real-time acquisition and output of raindrop image sequences for the target area from multiple perspectives via a multi-view imaging device. Real-time recording of the spatial position and temporal signal of raindrops crossing the laser plane via a laser array auxiliary module.

[0053] S2. Based on the image processing module, the raindrop image sequence of the target area and the spatial position and time signal of the raindrop crossing the laser plane are processed to identify and match the raindrop target and construct a three-dimensional raindrop trajectory.

[0054] S3. By combining the wind speed and direction data of the target area output by the particle parameter calculation module with the wind field information acquisition device, the constructed three-dimensional raindrop trajectory is dynamically analyzed, and the physical characteristic parameters of the raindrop are calculated and extracted.

[0055] S4. Through data processing and visualization modules, perform integrated calculations and visualize the dynamic three-dimensional raindrop trajectory, wind field superposition data, and the changing trends of raindrop physical property parameters over time and space.

[0056] It is understood that the 3D raindrop trajectory recognition method for wind-rain coupling scenarios in this embodiment can be divided into four core steps (steps S1 to S4), forming a complete closed loop from data acquisition to result presentation. The 3D raindrop trajectory recognition method for wind-rain coupling scenarios begins with the synchronous acquisition of multi-source heterogeneous data (step S1), followed by the core data processing and 3D reconstruction stage (step S2), then the analysis and extraction of physical parameters (step S3), and finally the data fusion and visualization output (step S4). These four steps are sequentially connected and progressively advance, with steps S2 and S3 incorporating wind field data as a core computational element into the processing flow, reflecting the core idea of ​​wind-rain coupling.

[0057] In the specific implementation process, each hardware module, under the synchronous signal scheduling of the central control module, first acquires wind vector data with strictly aligned timestamps, high-speed image sequences from multiple perspectives, and laser calibration signals in parallel. Subsequently, the image processing module preprocesses the image sequences, performs target detection and multi-target tracking, forming two-dimensional trajectory segments. During this process, wind field data is simultaneously used to predict motion trends to assist cross-view matching, while high-precision calibration is performed using absolute spatial coordinates provided by laser signals. Finally, a stereo vision algorithm fuses all information to generate an accurate three-dimensional trajectory. Next, the particle parameter calculation module uses the reconstructed three-dimensional trajectory and synchronized wind field data as input to calculate key physical characteristics of raindrops relative to the air, such as velocity, kinetic energy, and angle of incidence. Finally, all data is converged and fused, and a visualization engine dynamically renders a panoramic view of raindrop motion under the coupling effect of wind and rain, parameter change curves, etc., supporting interactive analysis and data export, thus transforming the complex physical process into intuitive and quantifiable analytical results.

[0058] In some embodiments of the three-dimensional raindrop trajectory recognition method applicable to wind and rain coupled scenarios of the present invention, the above step S2 may specifically include the following sub-steps S21 to S24.

[0059] S21. The image preprocessing submodule acquires multiple frames of raindrop images collected by the multi-view imaging device, and performs noise reduction, enhancement, and background extraction on the raindrop images.

[0060] S22. The target detection and segmentation submodule is used to identify raindrop targets in each preprocessed raindrop image and locate their spatial coordinates and time series.

[0061] S23. Based on the multi-target tracking submodule, the same raindrop is associated across frames in multiple raindrop images to form a two-dimensional trajectory segment of the raindrop.

[0062] S24. The stereo matching and three-dimensional reconstruction submodule uses the multi-view information from the multi-view imaging device and the wind speed and direction data output by the wind field information acquisition device. It combines the spatial position and time signal of the raindrop crossing the laser plane to fuse the formed two-dimensional trajectory segments of the raindrop, calculates the three-dimensional spatial coordinates of the raindrop, and constructs the three-dimensional raindrop trajectory.

[0063] It is understood that step S2 in this embodiment is subdivided into an automated processing pipeline consisting of four sub-steps (steps S21 to S24), which are progressively advanced. Structurally, this process begins with the image preprocessing submodule (step S21), followed by feature extraction by the target detection and segmentation submodule (step S22), then temporal correlation through the multi-target tracking submodule (step S23), and finally 3D information fusion by the stereo matching and 3D reconstruction submodule (step S24). Step S24 is not an independent visual computation process, but rather deeply integrates external physical information from the wind field and laser modules, forming a decision center for multi-source data fusion.

[0064] In practice, the image preprocessing submodule first performs filtering, contrast stretching, and background subtraction on the input multi-view original images to suppress noise and highlight moving raindrop targets. Then, the target detection and segmentation submodule (typically using a deep learning-based target detection network) analyzes each preprocessed frame, accurately identifying the contours of all raindrop targets and outputting their pixel coordinates and corresponding timestamps. Next, the multi-target tracking submodule applies a data association algorithm (such as Kalman filtering combined with the Hungarian algorithm) across frames to associate detection boxes belonging to the same raindrop in a time series, forming continuous two-dimensional motion trajectory segments of each raindrop from a single camera viewpoint. Finally, the stereo matching and 3D reconstruction submodule begins its work: receiving the tracked two-dimensional trajectory segments from different cameras and simultaneously introducing two key physical data streams—real-time wind speed and direction data and laser calibration signals (wind field data provides kinematic constraints for matching the same raindrop across different viewpoints, greatly improving the accuracy and robustness of matching in complex windy and rainy environments; while the laser signal provides a high-precision spatial absolute coordinate reference to correct systematic errors in the reconstruction algorithm). Ultimately, the stereo matching and 3D reconstruction submodule uses the principle of triangulation to integrate all information and calculate the precise coordinates of the raindrop in 3D space at every moment, thereby constructing a physically real and accurate 3D motion trajectory.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios, characterized in that, include: A wind field information acquisition device is installed in the target area to collect and output wind speed and wind direction data of the target area in real time. A multi-view imaging device is installed in the target area to acquire and output a sequence of raindrop images of the target area in real time from multiple perspectives. An image processing module is connected to the wind field information acquisition device and the multi-view imaging device. The image processing module preprocesses and segments the raindrop image sequence acquired by the multi-view imaging device, and introduces the real-time wind field data acquired by the wind field information acquisition device as a physical constraint to match and track raindrop targets between different views. The particle parameter calculation module is connected to the wind field information acquisition device and the image processing module. It is used to receive the three-dimensional raindrop trajectory data constructed by the image processing module, and calculate and extract the physical characteristic parameters of the raindrops based on the wind speed and wind direction data output by the wind field information acquisition device. The central control module is connected to the wind field information acquisition device, the multi-view imaging device, the image processing module, and the particle parameter calculation module respectively, for synchronous control and communication management. A laser array auxiliary module is disposed in the target area. The laser array auxiliary module is connected to the image processing module. The laser array auxiliary module generates a layered laser plane so that when raindrops pass through the layered laser plane, they produce bright spots with a brightness higher than the environmental background in the raindrop image sequence acquired by the multi-view imaging device. The spatial position of the bright spots corresponds to the known and precisely calibrated laser plane coordinates. The bright spots are used by the image processing module for recognition and extraction, and are used as a reference for spatial correction and calibration of the three-dimensional raindrop trajectory. The image processing module is configured to: preprocess, detect, and track multiple targets in the raindrop image sequence acquired by the multi-view imaging device to form a two-dimensional trajectory segment; simultaneously use the wind field data acquired by the wind field information acquisition device to predict the motion trend to assist cross-view matching; and use the coordinates of the raindrop high-brightness spot provided by the laser array auxiliary module for calibration; and fuse all information through a stereo vision algorithm to generate a three-dimensional trajectory of the raindrop.

2. The three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios according to claim 1, characterized in that, The laser array auxiliary module includes: Line laser generator, used to generate a static laser plane; A deflectable scanning galvanometer is disposed opposite to the line laser generator. The scanning galvanometer deflects the static laser plane generated by the line laser generator to form a layered spatial laser plane in the target area. The synchronization and control unit is connected to the central control module and the scanning galvanometer, respectively. The synchronization and control unit is used to receive the synchronization signal from the central control module and control the deflection angle and deflection speed of the scanning galvanometer.

3. The three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios according to claim 1, characterized in that, The three-dimensional raindrop trajectory recognition system suitable for wind and rain coupled scenarios also includes a data processing and visualization module, which includes: The data fusion submodule is connected to the wind field information acquisition device, the image processing module and the particle parameter calculation module, respectively, and is used to receive and fuse the wind speed and wind direction data, the physical characteristic parameters of raindrops and the three-dimensional raindrop trajectory of the target area. The physical parameter calculation engine, connected to the data fusion submodule, is used to calculate and generate three-dimensional raindrop trajectory dynamics, wind field superposition data, and the changing trends of raindrop physical property parameters over time and space. The visualization rendering engine, connected to the physical parameter calculation engine, is used to generate a three-dimensional dynamic map of raindrop trajectory, a wind field overlay map, and a map showing the changing trend of the physical characteristic parameters of each raindrop over time / space. The user interaction and output interface submodule includes a human-computer interaction interface and an output interface. The human-computer interaction interface is connected to the visualization rendering engine for visual graphic output; the output interface is connected to the physical parameter calculation engine for exporting data.

4. The three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios according to any one of claims 1 to 3, characterized in that, The wind field information acquisition device includes a three-dimensional ultrasonic anemometer or a wind vector sensor, which is used to acquire and output the wind speed and wind direction of the target area.

5. The three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios according to any one of claims 1 to 3, characterized in that, The multi-view imaging device includes at least two visual imaging devices and a background plate. The optical axes of the two visual imaging devices are intersected and intersect at the center of the target area. The background plate is located on the opposite side of the two visual imaging devices and serves as the imaging background for the two visual imaging devices. The background plate is provided with calibration scale.

6. The three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios according to claim 5, characterized in that, The multi-view imaging device also includes an LED lighting device located laterally to the target area, the LED lighting device being used to illuminate raindrops entering the target area from the side; The visual imaging device is a high-speed camera with a frame rate of not less than 5000fps.

7. The three-dimensional raindrop trajectory recognition system for wind and rain coupled scenarios according to any one of claims 1 to 3, characterized in that, The image processing module includes: The image preprocessing submodule is connected to the multi-view imaging device and is used to acquire multiple frames of raindrop images collected by the multi-view imaging device, and to perform noise reduction, enhancement and background extraction on the multiple frames of raindrop images. The target detection and segmentation submodule, connected to the image preprocessing submodule, is used to identify raindrop targets and locate their spatial coordinates and temporal sequence in each preprocessed raindrop image frame; The multi-target tracking submodule, connected to the target detection and segmentation submodule, is used to associate the same raindrop across frames in multiple raindrop images to form a two-dimensional trajectory segment of the raindrop; The stereo matching and 3D reconstruction submodule connects the wind field information acquisition device and the multi-target tracking submodule. Based on the multi-view information of the multi-view imaging device and the wind speed and wind direction data output by the wind field information acquisition device, the stereo matching and 3D reconstruction submodule fuses the formed two-dimensional trajectory segments of raindrops, calculates the three-dimensional spatial coordinates of the raindrops, and constructs the three-dimensional raindrop trajectory.

8. A method for recognizing three-dimensional raindrop trajectories in wind-rain coupled scenarios, applicable to the three-dimensional raindrop trajectory recognition system for wind-rain coupled scenarios as described in any one of claims 1 to 7, characterized in that, include: The wind speed and direction data of the target area are collected and output in real time through the wind field information acquisition device; Raindrop image sequences of the target area are acquired and output in real time from multiple perspectives using a multi-view imaging device; The spatial position and time signal of raindrops crossing the laser plane are recorded in real time by a laser array auxiliary module. Based on the image processing module, the raindrop image sequence of the target area and the spatial position and time signal of the raindrop crossing the laser plane are processed to identify and match the raindrop target and construct a three-dimensional raindrop trajectory; By combining the wind speed and direction data of the target area output by the particle parameter calculation module and the wind field information acquisition device, the constructed three-dimensional raindrop trajectory is dynamically analyzed, and the physical characteristic parameters of the raindrop are calculated and extracted. The data processing and visualization modules are integrated for calculation, and the dynamic three-dimensional raindrop trajectory, wind field superimposed data, and the changing trend of raindrop physical property parameters over time and space are visualized.

9. The three-dimensional raindrop trajectory recognition method for wind and rain coupled scenarios according to claim 8, characterized in that, The image processing module processes the raindrop image sequence of the target area and the spatial and temporal signals of raindrops crossing the laser plane to identify and match raindrop targets and construct a three-dimensional raindrop trajectory, including: The image preprocessing submodule acquires multiple frames of raindrop images captured by the multi-view imaging device, and performs noise reduction, enhancement, and background extraction on the raindrop images. The target detection and segmentation submodule is used to identify raindrop targets and locate their spatial coordinates and temporal sequence in each preprocessed raindrop image frame; Based on the multi-target tracking submodule, the same raindrop is associated across frames in multiple raindrop images to form a two-dimensional trajectory segment of the raindrop; The stereo matching and 3D reconstruction submodule uses multi-view information from a multi-view imaging device and wind speed and direction data from a wind field information acquisition device. It combines the spatial position and time signal of raindrops crossing the laser plane to fuse the formed two-dimensional trajectory segments of raindrops, calculate the three-dimensional spatial coordinates of the raindrops, and construct the three-dimensional raindrop trajectory.

Images