Wind speed vector experimental device and method based on multimode optical fiber speckle pattern monitoring

Abstract

This invention discloses an experimental device and method for wind speed vector monitoring based on multimode fiber speckle patterns, belonging to the field of fiber optic sensing technology. This invention solves the problems of existing wind measurement technologies, such as difficulty in adapting to the strong spatiotemporal heterogeneity of low-altitude wind fields, resulting in response lag, weak anti-interference, and low accuracy, and the inability to achieve precise vector monitoring of wind speed and direction. The device consists of a helium-neon laser, a 40x achromatic objective lens, multimode fiber, an octahedral frame, a CCD camera, and a computer connected sequentially. The multimode fiber is wound alternately in S-shape and Hexagonal shape around the octahedral frame. The experimental method involves calibrating the device under stable conditions, setting three wind speed levels from 0 to 4.5 m / s, collecting speckle patterns from multiple directions to construct a dataset of 24,000 images, and then training a deep learning model after segmentation. This invention divides the 0-4.5 m / s wind speed into three levels at 1.5 m / s intervals, achieving precise detection of 24 categories in eight directions. The classification accuracy is high, the device structure is simple, it is resistant to electromagnetic interference, and the detection real-time performance is good, laying the foundation for the large-scale application of fiber optic wind measurement technology.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing technology, and in particular to an experimental device and method for wind speed vector monitoring based on multimode fiber speckle patterns. Background Technology

[0002] With the rapid development of the low-altitude economy, applications such as drone logistics, low-altitude emergency rescue, and low-altitude security patrols have placed stringent and urgent demands on real-time, accurate, and multi-dimensional monitoring of low-altitude wind fields. Low-altitude areas are affected by factors such as building obstruction, terrain undulations, and the urban heat island effect, resulting in strong spatiotemporal heterogeneity in wind fields. Phenomena such as localized eddies and sudden wind speed changes occur frequently. Traditional meteorological forecasting systems, due to insufficient spatial resolution, cannot accurately depict the dynamic changes of such microscale wind fields, becoming a core hidden danger to the safe operation of low-altitude operations and severely restricting the large-scale development of the low-altitude economy.

[0003] Currently, mechanical, acoustic, and thermal anemometers are the mainstream monitoring methods in wind measurement technology. However, these devices generally suffer from drawbacks such as slow response, weak resistance to environmental interference, and high power consumption. Furthermore, they are mostly single-point, single-direction monitoring modes, making it difficult to meet the systematic and networked deployment requirements for low-altitude wind field monitoring. Fiber optic anemometers, with their advantages of resistance to electromagnetic interference and harsh environments, have become an important development direction for wind measurement technology. However, existing fiber optic anemometers are mostly based on fiber grating or interferometric structure designs. Their measurement principle relies on a fixed physical model for wind speed inversion, making them highly susceptible to measurement drift caused by external factors such as temperature, vibration, and fiber micro-bending. Frequent calibration and compensation are required, resulting in insufficient system robustness and an inability to adapt to the complex and ever-changing low-altitude monitoring environment.

[0004] Multimode fiber speckle sensing technology has gained widespread attention in the sensing field in recent years due to its extremely high sensitivity to weak external disturbances such as deformation and airflow disturbances. Furthermore, it requires no complex physical tuning models, boasts a simple structure, and is inexpensive to manufacture. The core principle of this technology is that the guided modes within the multimode fiber interfere to form a characteristic speckle pattern. When an external disturbance occurs, the fiber undergoes a slight deformation, leading to changes in the transmission phase and optical path of the guided modes. This results in a regular change in the speckle pattern, and the external disturbance can be detected by analyzing this change. Currently, this technology has been successfully applied to the monitoring of deformation, touch, and pressure. However, its application in low-altitude wind speed vector monitoring remains in the exploratory stage, and a mature application solution has not yet been developed. Existing research largely remains at the laboratory verification stage, failing to address the core characteristics of strong spatiotemporal heterogeneity and multidirectional airflow disturbances in low-altitude wind fields through sensor structure design. It also lacks integration with intelligent analysis technology to achieve accurate vector monitoring of wind speed and direction, making it difficult to meet the real-time, accurate, and stable requirements of low-altitude economic wind speed monitoring.

[0005] Therefore, developing a wind measurement device that can adapt to complex low-altitude wind field environments, achieve high-precision monitoring of wind speed vectors, and is simple in structure, has strong anti-interference capabilities, and is easy to deploy on a large scale has become a key technological requirement for promoting the economic and safe development of low-altitude areas. Summary of the Invention

[0006] The purpose of this application is to overcome the problems in the prior art and provide an experimental device and method for wind speed vector monitoring based on multimode fiber speckle pattern.

[0007] The technical concept of this application revolves around the need for precise monitoring of low-altitude wind fields. It integrates multimode fiber speckle sensing technology with deep learning algorithms, and designs an octahedral frame carrying a step-index multimode fiber sensing structure. The fiber is wound around the eight faces of the frame in an alternating S-shape and Hexagonal pattern. A helium-neon laser is used to couple the laser to excite the guided mode interference within the fiber to form a speckle pattern. A CCD camera is used to acquire speckle images under airflow disturbances at different wind speeds and directions. A standardized dataset is constructed and divided into training, validation, and test sets according to proportions. A deep learning model is used to extract speckle features and complete wind speed vector classification. By utilizing the changes in speckle pattern caused by the micro-deformation of the fiber due to airflow disturbances, precise detection of 24 categories in eight directions with wind speeds of 0-4.5 m / s is achieved. This approach combines the advantages of simple structure, resistance to electromagnetic interference, and low cost, and is suitable for the systematic and networked monitoring needs of complex low-altitude wind fields.

[0008] To achieve the aforementioned objectives, the present invention employs the following technical solution: a wind speed vector experimental device based on multimode fiber speckle pattern monitoring, comprising a helium-neon laser, a 40x achromatic objective lens, an optical fiber pigtail, a multimode optical fiber, a collimating lens, a CCD camera, a USB data cable, and a computer connected in sequence, and also including an octahedral frame; the multimode optical fiber is wound and fixed on the columns of the octahedral frame, and the computer has built-in image acquisition software and a Python data processing environment, and is equipped with a deep learning algorithm.

[0009] Furthermore, the octahedral frame is a regular octagonal hollow aluminum alloy frame with a side length of 15cm and a height of 75cm. The multimode optical fiber is wound around the eight pillars of the octahedral frame in an alternating S-shaped and He-shaped pattern, forming an optical fiber winding surface between adjacent pillars. The winding methods of adjacent optical fiber winding surfaces are different, and the multimode optical fibers are taut and evenly distributed on each face. This configuration can expand the airflow sensing range of the sensing unit, improve the sensitivity to airflow disturbances in different directions, and make it easier for lasers to couple into the multimode optical fiber and excite multiple guided modes, laying a structural foundation for accurate wind speed vector detection.

[0010] Furthermore, the multimode fiber is a step-index multimode fiber with a core diameter of 50 μm, a cladding diameter of 125 μm, and a numerical aperture of 0.22. This specification of multimode fiber ensures the stability of guided-mode interference, makes the fiber highly sensitive to micro-deformations caused by airflow disturbances, ensures that the speckle pattern exhibits regular characteristic changes with wind speed, and improves the accuracy of detection.

[0011] Furthermore, the helium-neon laser and the 40x achromatic objective lens are fixed together on the adjustment frame. The laser, focused by the 40x achromatic objective lens, is aligned and docked with the fiber optic pigtail to achieve efficient coupling of the laser to the multimode fiber. The output end of the multimode fiber is coaxially docked with the collimating lens, and the speckle beam is collimated by the collimating lens to form a parallel beam and projected onto the target surface of the CCD camera. This eliminates the imaging blurring problem caused by beam divergence, allowing the speckle pattern to be acquired clearly and completely.

[0012] Furthermore, the CCD camera has a resolution of 1368×918 pixels, enabling real-time, lossless transmission of speckle pattern data to the computer via a USB data cable. This resolution fully preserves the core features of the speckle pattern, such as grayscale, texture, and edges, while real-time, lossless transmission ensures the timeliness and integrity of data acquisition, providing high-quality raw data for subsequent deep learning model training.

[0013] An experimental method for a wind speed vector experimental device based on multimode fiber speckle pattern monitoring includes the following steps: S1. Place the wind speed vector experimental device in a stable environment with a temperature of 15-25℃ and a humidity of 40-50% RH, and complete the calibration and connection debugging of each component of the device; avoid the interference of temperature, humidity and stray airflow on the measurement results, and ensure the stability of the device operation.

[0014] S2. The wind speed of 0-4.5m / s is divided into three target wind speed levels, from level 1 to level 3, in 1.5m / s intervals. An adjustable speed fan is used to provide airflow, and a handheld split-type anemometer with a resolution of 0.01m / s is used to complete the wind speed calibration. This achieves precise control of wind speed in different levels and provides a standard for the classification and acquisition of speckle patterns.

[0015] S3. Collect speckle patterns of different wind speeds and different faces of the octahedral frame using a CCD camera, and transfer them to a computer for storage via a USB data cable; to achieve comprehensive acquisition of speckle data at multiple speeds and in multiple directions.

[0016] S4. The computer-based system segments the collected speckle pattern dataset and trains a deep learning model to classify and detect wind speed vectors. Deep learning algorithms are used to mine the correlation between speckle features and wind speed vectors, achieving accurate classification and recognition.

[0017] Further, in step S2, the specific operation of the wind speed calibration is as follows: Position the adjustable-speed fan directly opposite one of the fiber optic winding surfaces of the octahedral frame. Place a handheld split-type anemometer at three measuring points on the upper, middle, and lower sides of this fiber optic winding surface. Adjust the distance between the fan and the frame so that the wind speed at all three measuring points falls within the same target wind speed level. Record the corresponding distances and complete the calibration for the three levels sequentially. This multi-measuring-point calibration method ensures the uniformity of the wind speed received by a single fiber optic winding surface, avoiding local wind speed deviations from affecting the accuracy of speckle pattern acquisition, and providing assurance for the reliability of subsequent experimental data.

[0018] Further, in step S3, the specific operation of speckle pattern acquisition is as follows: The adjustable-speed fan is placed at the calibrated corresponding distance, and the CCD camera is activated to continuously acquire 100 valid speckle images of 1368×918 pixels at the current wind speed setting. After completion, the acquisition is repeated at the next setting, sequentially acquiring speckle patterns on all eight faces of the octahedral frame. A total of 10 repeated measurements are performed, resulting in 24,000 speckle patterns forming the original dataset. This large-sample original dataset enriches the coverage of speckle features, reduces the impact of random factors on model training, and improves the accuracy and generalization ability of the deep learning model for wind speed vector classification.

[0019] Furthermore, the acquisition parameters of the CCD camera are set to: a frame rate of 2fps and an exposure time of 20000μs, ensuring that the acquired speckle pattern fully retains grayscale, texture, and edge core features. These acquisition parameters are adapted to the camera resolution, maximizing the preservation of key features of the speckle pattern while ensuring acquisition efficiency. This allows the deep learning model to accurately extract feature information corresponding to wind speed and direction, improving the accuracy of classification and detection.

[0020] Further, in step S4, the original dataset is divided into a training set, a validation set, and a test set in a 7:1.5:1.5 ratio. These sets are used for training the deep learning model, optimizing hyperparameters, and testing generalization ability, respectively, resulting in a confusion matrix and a validation accuracy map. This dataset division ensures sufficient training for the model. Simultaneously, the validation set optimizes hyperparameters, and the test set verifies generalization ability, enabling the trained model to possess stable wind speed vector classification and detection capabilities. The confusion matrix and validation accuracy map also visually reflect the model's classification performance, providing a basis for further model optimization.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention innovatively integrates multimode fiber speckle patterns with deep learning algorithms to create an integrated intelligent wind speed vector experimental device, achieving a highly efficient combination of sensing structure and intelligent analysis. By designing an alternating S-shaped and He-shaped winding method for the multimode fiber on an eight-sided hollow aluminum alloy frame, the helium-neon laser can be coupled into the interior of the multimode fiber more efficiently, fully exciting multiple guided modes. This allows the light to complete stable transmission interference in the multimode fiber and form a characteristic speckle pattern, laying a solid structural foundation for the accurate detection of wind speed vectors. Simultaneously, the integrated structural design allows for tighter connections between the various components, improving the stability of light transmission and speckle acquisition, and making the overall device easier to assemble and debug, significantly reducing the complexity of experimental operations and providing structural support for subsequent large-scale applications.

[0022] 2. This invention fully utilizes the guided-mode interference characteristics of multimode optical fibers to achieve high-sensitivity and rapid-response detection of wind speed vectors, solving the technical pain point of slow response in traditional wind measurement devices. A large number of guided modes are formed inside the multimode optical fiber due to laser excitation. When the external wind speed changes, airflow disturbances directly cause minute deformations in the fiber, thereby altering the transmission phase and optical path of the guided modes, resulting in a regular and corresponding change in the speckle pattern. Based on this, a deep learning model is used to intelligently analyze and accurately identify the features of the speckle pattern, quickly capturing the correlation between speckle changes and wind speed and direction. This enables real-time and accurate detection of the wind environment, with a fast and stable detection response. It effectively adapts to the complex characteristics of sudden wind speed changes and multi-directional disturbances in low-altitude wind fields, improving the timeliness and accuracy of wind measurement.

[0023] 3. This invention pioneers a technical solution for monitoring wind speed vectors using multimode fiber speckle patterns, filling a gap in related technological applications and possessing numerous practical advantages. It opens a new path for the research and large-scale application of high-precision wind speed vector sensors. This technology achieves efficient monitoring of wind speed vectors through a simple fiber winding method and automated data acquisition, significantly improving wind speed vector identification accuracy and scalability. It provides a novel method for fabricating high-precision wind speed vector sensors that are simple in structure, small in size, and low in cost. Experiments verify that it classifies wind speeds from 0-4.5 m / s into 3 levels and eight directions to form 24 categories, achieving a classification accuracy of up to 99.86%. Compared to other fiber optic wind speed sensors, it has significant advantages in high accuracy, simple manufacturing, and strong anti-interference capabilities, greatly improving practicality and adaptability, and laying a solid foundation for the large-scale application of fiber optic wind measurement technology. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0025] Figure 1 This is a schematic diagram of the wind speed vector sensing experimental device in this invention. Figure 1 .

[0026] Figure 2 This is a schematic diagram of the wind speed vector sensing and measurement experimental device in this invention. Figure 2 .

[0027] Figure 3 Figure 1 shows a side view of the octahedral frame in this invention, where Figure (a) is a schematic diagram of the S-shaped optical fiber winding surface and Figure (b) is a schematic diagram of the He-shaped optical fiber winding surface.

[0028] Figure 4 This is the confusion matrix for wind speed vector classification in this invention.

[0029] Figure 5 This is a graph showing the accuracy of the verification of the present invention.

[0030] Figure 6 Software parameter settings and image acquisition during the experiment of this invention.

[0031] The attached figures are labeled as follows: 1. Helium-neon laser; 2. 40x achromatic objective lens; 3. Fiber optic pigtail; 4. Multimode fiber; 5. Octahedral frame; 6. Collimating lens; 7. CCD camera; 8. USB data cable; 9. Computer; 10. Adjustable speed fan; 11. Handheld split-type anemometer. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0033] This embodiment details the structural composition of the wind speed vector experimental device based on multimode fiber speckle pattern monitoring and the corresponding wind speed detection experimental method. The experimental device and method of this invention are suitable for high-precision monitoring of low-altitude wind speed vectors from 0 to 4.5 m / s, and can achieve accurate classification and detection of 24 categories in 8 directions and 3 wind speed levels. The following is a detailed description of this embodiment with reference to the accompanying drawings.

[0034] like Figure 1 and Figure 2As shown, the wind speed vector experimental device based on multimode fiber speckle pattern monitoring of the present invention mainly consists of a helium-neon laser 1, a 40x achromatic objective lens 2, an optical fiber pigtail 3, a multimode fiber 4, an octahedral frame 5, a collimating lens 6, a CCD camera 7, a USB data cable 8, and a computer 9 connected sequentially. These components work together to achieve laser coupling, speckle formation, image acquisition, and intelligent data analysis. The overall structure is compact and highly functional. The helium-neon laser 1 and the 40x achromatic objective lens 2 are fixed together on an adjustment frame, allowing for fine adjustment of the laser emission angle to ensure that the laser is collimated and parallel to the center of the objective lens. The multimode fiber 4 is a step-index fiber with a core diameter of 50 μm, a cladding diameter of 125 μm, and a numerical aperture of 0.22. These specifications ensure the stability of guided-mode interference and high sensitivity to airflow disturbances. The octahedral frame 5 is a regular octagonal hollow aluminum alloy frame with a side length of 15cm and a height of 75cm. The multimode optical fiber 4 is taut and wound around the columns on the eight sides of the frame in an alternating S-shaped and He-shaped pattern. The winding patterns on adjacent sides are different and the optical fibers are evenly distributed, as shown in Figure 3, which effectively expands the airflow sensing range. The input end of the multimode optical fiber 4 is connected to the laser calibration after being focused by the objective lens via the fiber optic pigtail 3 to achieve efficient laser coupling. The output end is coaxially connected to the collimating lens 6, which can calibrate the speckle beam into a parallel beam. The CCD camera 7 has a resolution of 1368×918 pixels and is set directly in front of the collimating lens 6. It is stably connected to the computer 9 via the USB data cable 8. The computer 9 has built-in image acquisition software, a Python data processing environment, and is equipped with deep learning algorithms, which can complete the real-time storage, preprocessing, feature extraction, and wind speed vector classification of speckle patterns.

[0035] II. Wind Speed ​​Detection Experimental Method The wind speed detection experiment based on the above experimental setup strictly follows the steps of environmental control, setup debugging, wind speed calibration, speckle data acquisition, and model training. The specific operation is as follows: Step 1: Experimental Environment and Apparatus Debugging: Place the assembled experimental apparatus in a stable environment with a temperature of 15-25℃ and a humidity of 40-50% RH to avoid interference from temperature and humidity fluctuations and stray airflow on the measurement results.

[0036] First, adjust the mounting brackets of the helium-neon laser 1 and the 40x achromatic objective lens 2, adjusting the horizontal, vertical, and pitch angles to ensure the laser beam is collimated and parallel, and precisely incident on the center of the objective lens, guaranteeing optimal focusing. Next, perform a secondary calibration and docking of the focused laser beam with the fiber optic pigtail 3, fine-tuning the relative position of the pigtail and the focused spot to achieve efficient coupling of the laser beam to the multimode fiber 4, reducing energy loss. Then, adjust the relative position of the output end of the multimode fiber 4 and the collimating lens 6, so that the speckle beam is formed into a clear parallel beam by the collimating lens 6 and completely projected onto the target surface of the CCD camera 7. Finally, connect the CCD camera 7 to the computer 9 via the USB data cable 8, open the accompanying image acquisition software on the computer, preview the speckle pattern, and confirm that the image is clear, without ghosting, and without noise, completing the overall setup and debugging of the device.

[0037] Step 2, Wind speed level division and calibration: Divide the wind speed of 0-4.5m / s into 3 target wind speed levels at 1.5m / s intervals, namely Level 1 0-1.5m / s, Level 2 1.5-3m / s, and Level 3 3-4.5m / s. An adjustable-speed fan 10 was used to provide the experimental airflow. A handheld split-type anemometer 11 with a resolution of 0.01 m / s was used to calibrate the wind speed. The adjustable-speed fan 10 was positioned facing one of the fiber optic winding surfaces of the octahedral frame 5. The anemometer 11 was placed at three measuring points on the winding surface: the upper, middle, and lower. The fan was turned on to level 1, and the distance between the fan and the frame was slowly adjusted. The wind speed at the three measuring points was monitored in real time until the wind speed at all measuring points was stably within the level 1 range. The distance D1 between the fan and the frame at this time was recorded. Similarly, the fan was adjusted to level 2 and level 3, and the corresponding distances D2 and D3 were calibrated and recorded respectively. After the wind speed calibration of a single winding surface was completed, the remaining seven winding surfaces were calibrated in the same way to ensure that the wind speed of each surface and each level was accurate and controllable.

[0038] Step 3: Speckle Pattern Acquisition and Dataset Construction: At the calibrated distance D1, position the adjustable fan 10 directly opposite the first winding surface of the octahedral frame 5, set it to level 1, and maintain stable airflow. Start the CCD camera 7, setting the acquisition parameters to a frame rate of 2fps and an exposure time of 20000μs. Continuously acquire 100 valid 1368×918 pixel speckle images at the current level. After acquisition, switch the fan to level 2 at distance D2 and repeat the acquisition operation. Then switch to level 3 at distance D3 and repeat the acquisition operation to complete the acquisition. After acquiring speckle patterns at all three levels for a single winding surface, sequentially acquire speckle patterns at each level for the remaining seven winding surfaces of the frame using the same method. This constitutes one complete acquisition. A total of 10 repeated measurements are performed, ultimately obtaining 24,000 speckle patterns to construct the original experimental dataset. During the acquisition process, ensure stable laser output and clear CCD camera imaging to ensure that the speckle patterns fully retain core features such as grayscale, texture, and edges.

[0039] Step 4: Dataset Partitioning and Deep Learning Model Training: The original dataset of 24,000 speckle patterns was divided into training, validation, and test sets in a 7:1.5:1.5 ratio. The training set was used for feature learning and training of the deep learning model, the validation set was used for optimizing and adjusting the model's hyperparameters, and the test set was used to verify the model's generalization ability. In the Python data processing environment of PC9, the speckle patterns in each dataset were preprocessed to extract key features such as grayscale, texture, and edges. The processed feature data was then input into the deep learning model for training. During training, the model's hyperparameters were continuously optimized using the validation set to improve the model's recognition accuracy. After training, the model's performance was tested using the test set, generating a corresponding confusion matrix to visually reflect the model's classification performance for the 24 wind speed vector categories.

[0040] The experimental setup in this embodiment has a simple structure, strong resistance to electromagnetic interference, and the experimental methods are standardized and the data is highly reliable. Figure 4-6 As shown in the test, the device and method achieved a classification accuracy of up to 99.86% for 24 categories in 8 directions for wind speeds of 0-4.5 m / s. It can realize real-time and accurate monitoring of low-altitude wind speed vectors, effectively solving the problems of slow response, weak anti-interference and low accuracy of traditional wind measurement technology. It provides reliable experimental basis and technical support for the large-scale application of fiber optic wind measurement technology.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A wind speed vector experimental device based on multimode optical fiber speckle pattern monitoring, characterized in that, The system includes a helium-neon laser (1), a 40x achromatic objective lens (2), a fiber optic pigtail (3), a multimode fiber (4), a collimating lens (6), a CCD camera (7), a USB data cable (8), and a computer (9) connected in sequence. It also includes an octahedral frame (5). The multimode fiber (4) is wound and fixed on the column of the octahedral frame (5). The computer (9) has built-in image acquisition software and a Python data processing environment, and is equipped with a deep learning algorithm. 2.The experimental device for monitoring wind speed vector based on speckle pattern of multimode fiber according to claim 1, wherein, The octahedral frame (5) is a regular octagonal hollow aluminum alloy frame; the multimode optical fiber (4) is wound on the eight pillars of the octahedral frame (5) in an alternating S-shape and J-shape, and an optical fiber winding surface is formed between two adjacent pillars, and the winding methods of two adjacent optical fiber winding surfaces are different. The multimode optical fiber (4) is straight and evenly distributed on each side. 3.The experimental device for wind speed vector based on multi-mode fiber speckle pattern monitoring according to claim 1, wherein, The multimode fiber (4) is a step-index multimode fiber with a core diameter of 50 μm, a cladding diameter of 125 μm, and a numerical aperture of 0.

22. 4.The experimental device for wind speed vector based on multi-mode fiber speckle pattern monitoring according to claim 1, wherein, The helium-neon laser (1) and the 40x achromatic objective (2) are fixed together on the adjustment frame. The laser, after being focused by the 40x achromatic objective (2), is aligned and docked with the fiber optic pigtail (3) to achieve efficient coupling of the laser to the multimode fiber (4). The output end of the multimode fiber (4) is coaxially docked with the collimating lens (6). The speckle beam is collimated by the collimating lens (6) to form a parallel beam and is projected onto the target surface of the CCD camera (7).

5. The experimental device for wind speed vector monitoring based on multi-mode optical fiber speckle pattern according to claim 1, characterized in that, The CCD camera (7) has a resolution of 1368×918 pixels and transmits speckle pattern data to the computer (9) in real time without loss via a USB data cable (8).

6. An experimental method of a wind speed vector experimental device based on the multimode fiber speckle pattern monitoring according to any one of claims 1-5, characterized in that, Includes the following steps: S1. Place the wind speed vector experimental device in a stable environment with a temperature of 15-25℃ and a humidity of 40-50% RH, and complete the calibration and connection debugging of each component of the device; S2. Divide the wind speed of 0-4.5m / s into three target wind speed levels, from level 1 to level 3, at intervals of 1.5m / s. Use an adjustable speed fan (10) to provide airflow and complete the wind speed calibration with a handheld split-type anemometer (11) with a resolution of 0.01m / s. S3. Collect speckle patterns of different faces of the octahedral frame (5) at different wind speeds using a CCD camera (7), and transmit them to a computer (9) via a USB data cable (8) for storage. S4. The computer (9) terminal divides the collected speckle pattern dataset and trains a deep learning model to achieve classification and detection of wind speed vectors.

7. The experimental method of claim 6, wherein, In step S2, the specific operation of the wind speed calibration is as follows: the adjustable speed fan (10) is placed facing one of the fiber winding surfaces of the octahedral frame (5), and the handheld split anemometer (11) is placed at the upper, middle and lower measuring points of the fiber winding surface. The distance between the adjustable fan (10) and the frame is adjusted so that the wind speed at the three measuring points falls into the same target wind speed level. The corresponding distance is recorded and the calibration of the three levels is completed in sequence.

8. The experimental method of claim 6, wherein, In step S3, the specific operation of speckle pattern acquisition is as follows: the adjustable speed fan (10) is placed at the corresponding distance after calibration, the CCD camera (7) is started to continuously acquire 100 valid speckle patterns of 1368×918 pixels at the current wind speed level, and after completion, the acquisition is repeated at the next speed level. The speckle patterns of the eight faces of the octahedral frame (5) are acquired in sequence, and a total of 10 repeated measurements are carried out to obtain 24,000 speckle patterns to form the original dataset.

9. The experimental method of claim 8, wherein, The acquisition parameters of the CCD camera (7) are set as follows: frame rate 2fps, exposure time 20000μs, and the acquired speckle pattern fully retains grayscale, texture, and edge core features.

10. The experimental method according to claim 8, characterized in that, In step S4, the original dataset is divided into a training set, a validation set, and a test set in a ratio of 7:1.5:1.5, which are used for training the deep learning model, optimizing hyperparameters, and testing generalization ability, respectively, to obtain the corresponding confusion matrix.

Images