Fan type rapid identification method for maritime work laboratory motion measurement

Abstract

The invention belongs to the technical field of ocean engineering experiments, and provides a fan type rapid identification method for maritime work laboratory motion measurement, which relates to a device comprising a camera for motion measurement, an image collector and a PC (Personal Computer). According to the method, superpixel analysis and optical flow analysis are adopted, K-means clustering is combined to obtain the area of the blade on the image, and in order to reduce the error rate of discrimination, the fan category is discriminated on the image according to the layout characteristics of the blade, and the fan category is judged again through the mutual relation of dynamic motion vectors. According to the method, the type of the floating fan model can be rapidly identified, the type of the measured target can be rapidly obtained at the beginning of motion measurement, the influence on the operation load in the motion measurement process is small, integration and intellectualization promotion of a whole fan experiment test system are facilitated, and the method has high practical value.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of ocean engineering experiments, and particularly relates to a fan type rapid identification method for motion measurement in a marine engineering laboratory. BACKGROUND

[0002] New energy technology, as an important part of the new technological revolution, not only can alleviate the growing demand for electricity, but also helps to solve the power supply problem of remote islands. The floating wind turbine that can be arranged in the deep sea is one of the widely concerned power generation forms in new energy technology.

[0003] Due to the difficulty and high cost of ocean engineering implementation, in order to avoid huge economic losses caused by design errors, the designed fan is often verified in the wave pool of the ocean engineering laboratory. In this process, a fan model needs to be constructed according to a certain scale ratio for experimental testing, including measuring the force load of the fan and the motion of the fan model. In terms of measuring the motion of the model, in order to reduce the influence of the added sensor on the motion of the model, visual measurement technology is often used for non-contact motion measurement in the laboratory.

[0004] Currently, when using visual measurement technology to measure the fan model, the user often needs to first confirm the type of the test fan, that is, to distinguish whether it is a horizontal axis fan or a vertical axis fan, and then select different image processing procedures according to the test characteristics of horizontal axis and vertical axis fans (because the motion projection of different types of fans on the image plane is different) to measure the motion of the fan blades and the fan foundation. In the fan type confirmation stage, in the traditional experiment, the user software setting method is generally used, and now the automatic recognition method is often used, such as using deep neural network for learning and training and recognition, to improve the automation and intelligence level of measurement, thereby simplifying the operation complexity of the experimental personnel. However, in the process of fan recognition using deep learning, it is necessary to continuously increase the learning samples to correctly identify the fan. However, new fan forms are constantly emerging in scientific research frontiers, and when a new fan design is developed, it is often not correctly identified if it is not included in the training samples, so it is not possible to form a stable measurement system for delivery to users, which is not conducive to the promotion of measurement technology. Therefore, it is of great practical value to construct a fan type rapid identification method for motion measurement in a marine engineering laboratory. SUMMARY

[0005] Aiming at the problem of fan model experiment process fan type automatic identification, the application carries out deep mining to fan model motion measurement characteristics, uses the camera used in visual measurement as image acquisition equipment, and on the basis of digital image processing technology, proposes a fan type fast identification method used in marine laboratory motion measurement. The method uses the characteristics of horizontal axis fan and vertical axis fan motion mode to automatically identify the motion mode, and the execution process uses the image algorithm with low operation complexity and mature application to improve the identification efficiency, and the result of the optical flow analysis can also be used as the basis for the next stage motion tracking, which is convenient for integration with the visual measurement system. Since the application does not involve sample learning training process, the type of fan model can be quickly identified by only several traditional image processing operators, so that the usability and universality of the technology are greatly improved.

[0006] The technical scheme of the application:

[0007] A fan type fast identification method used in marine laboratory motion measurement, the used device includes a camera for motion measurement, an image collector and a PC; the camera is connected with the image collector, and the image collector is connected with the PC. The arrangement of the camera is set according to the motion range of the fan model, so that the fan model is always in the range of the camera view area. When identifying the fan type, the image collected by the camera is subjected to super pixel image analysis and optical flow analysis, then the region of the fan blade on the image is obtained through K-means clustering and target selection, and finally the type of the fan model is identified through static feature analysis and dynamic motion analysis.

[0008] The specific steps are as follows:

[0009] Step A: arranging the fan model in the experimental pool, installing the camera for measurement in front of the fan model and requiring the view area to cover the entire motion range of the fan model, then the image acquired by the camera is collected by the image collector and transmitted to the user PC;

[0010] Step B: using the image super pixel method to perform image segmentation on the collected image, the image is changed from pixel to super pixel region, and the optical flow method is used to analyze the motion of the images at the front and rear time;

[0011] Step C: according to the judgment criterion that the same rigid body has continuous motion vector, using K-means clustering algorithm to cluster each super pixel to form different target regions;

[0012] Step D: on the basis of the result of step C, selecting the long strip target with the aspect ratio greater than the threshold and meeting the region size exceeding the threshold The length direction is calculated, and the average motion vector of the center of the selected elongated target, which is the blade, is calculated based on the optical flow analysis results.

[0013] Step E: When the similarity along the length direction of the elongated target is greater than the threshold If the similarity of the average motion vector direction at the center of all elongated targets is greater than a threshold, it is judged as a vertical axis fan; if the length direction of the elongated target has a certain angle and the average motion vector direction at the center of the elongated target has a certain angle with each other, it is judged as a horizontal axis fan; other states cannot be distinguished.

[0014] Step F: If the current analysis fails to identify the type, repeat steps B-E until identification is successful. Once successful, stop the above type identification process and begin the motion measurement phase.

[0015] The beneficial effects of this invention are as follows: In the process of identifying wind turbine model types, the contradiction between the heavy reliance of learning-based methods on training samples and the continuous innovation of wind turbine shapes in scientific research experiments is fully considered. At the same time, the universality and speed of the computational method, as well as the ease of integration with the motion measurement system, are also emphasized. Therefore, in the design process, two relatively mature methods, superpixel analysis and optical flow analysis, are adopted, combined with K-means clustering to obtain the blade regions on the image. Next, to reduce the error rate, the wind turbine category is not only identified based on the layout features of the blades in the image, but also further determined through the interrelationship of dynamic motion vectors. This invention achieves rapid category identification of floating wind turbine models, not only quickly obtaining the category of the target at the beginning of motion measurement, but also having a small impact on the computational load of the motion measurement process. It facilitates the integration and intelligent advancement of the entire wind turbine experimental testing system and has high practical value. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the apparatus used in the method of the present invention.

[0017] Figure 2 This is a flowchart illustrating the method of the present invention.

[0018] In the diagram: 1. Camera; 2. Image acquisition device; 3. PC; 4. Horizontal axis fan; 5. Vertical axis fan; 6. Fan blades. Detailed Implementation

[0019] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0020] A rapid identification method for wind turbine type in motion measurement in marine engineering laboratories, using the following device: Figure 1As shown, the system includes a camera 1 for motion measurement, an image acquisition unit 2, and a PC 3. Camera 1 is connected to image acquisition unit 2 via a dedicated cable, while image acquisition unit 2 (which can be an external image acquisition box or a PCIe bus image acquisition card) is connected to PC 3 via a dedicated interface. The camera 1 is arranged according to the requirements of motion measurement, specifically based on the movement range of the horizontal axis fan 4 or the vertical axis fan 5, ensuring that the horizontal axis fan 4 or the vertical axis fan 5 is always within the field of view of camera 1. After the equipment is installed, when identifying the fan type, the image undergoes superpixel image analysis and simultaneous optical flow analysis. Then, K-means clustering and target selection are used to obtain the region of blade 6 in the image. Finally, static feature analysis and dynamic motion analysis are used to identify the type of horizontal axis fan 4 or vertical axis fan 5. The specific method is as follows... Figure 2 As shown, the description is as follows:

[0021] Step A: Arrange the horizontal axis fan 4 or the vertical axis fan 5 in the experimental water tank, and install the measuring camera 1 in front of the horizontal axis fan 4 or the vertical axis fan 5, ensuring that the field of view can cover the entire range of motion of the horizontal axis fan 4 or the vertical axis fan 5; then, the image will be transmitted to the user's PC 3 through the image acquisition device 2 (which can be an external dedicated interface image acquisition box, or a PCIE image acquisition card inserted into the PC);

[0022] Step B: Use superpixelation to segment the acquired images, and use optical flow to perform motion analysis on the images at different time points.

[0023] Step C: Based on the criterion that the same rigid body has continuous motion vectors, the K-means clustering algorithm is used to cluster each superpixel to form different target regions;

[0024] Step D: Based on the results of Step C, select elongated targets (aspect ratio 1). , For set values, such as And the region size exceeds a certain threshold ( ,in For set values, such as The length direction is calculated, and the average motion vector of the center of the selected elongated target (blade) is calculated based on the optical flow analysis results. (Normally, what is elongated may not be a blade or may be a tower column, but after clustering in step 2, the elongated shape must be a blade because the marine engineering test scenario is relatively simple).

[0025] Step E: When the similarity in the length direction of the elongated target (wind turbine blade 6) ( For set values, such as Furthermore, the average motion vector direction similarity at the center of all 6 blades is... ( For set values, such as When calculating, the vector direction is taken as the absolute value, that is, regardless of whether the directions are exactly the same or completely opposite, they are considered to be consistent. It is judged as a vertical axis fan 5; when the length direction of the long strip target (fan blade 6) has a certain angle (length direction similarity) When the average motion vector directions at the center of the blade are at a certain angle to each other (motion vector direction similarity) The first one is identified as horizontal axis fan 4; the other states cannot be identified.

[0026] Step F: If the current analysis fails to identify the type, repeat steps B-E until identification is successful. Once successful, stop the above type identification process and begin the motion measurement phase.

Claims

1. A method for rapid identification of wind turbine type in motion measurement in marine engineering laboratories, characterized in that, The apparatus used includes a camera for motion measurement, an image acquisition unit, and a PC. The camera is connected to the image acquisition unit, which in turn is connected to the PC. The camera is positioned according to the motion range of the wind turbine model to ensure that the wind turbine model is always within the camera's field of view. When identifying the wind turbine type, superpixel image analysis and optical flow analysis are performed on the images acquired by the camera. Then, K-means clustering and target selection are used to obtain the region of the wind turbine blades in the image. Finally, static feature analysis and dynamic motion analysis are used to identify the type of wind turbine model.

2. The method for rapid identification of wind turbine type in motion measurement in marine engineering laboratories according to claim 1, characterized in that, The specific steps are as follows: Step A: Set up the wind turbine model in the experimental water tank, and install a measuring camera in front of the wind turbine model, ensuring that the field of view covers the entire range of motion of the wind turbine model; then, the images acquired by the camera will be collected by an image acquisition device and transmitted to the user's PC. Step B: Use the superpixel method to segment the acquired image, transforming the image from pixels into superpixel regions. At the same time, use optical flow to perform motion analysis on the images at different time points. Step C: Based on the criterion that the same rigid body has continuous motion vectors, the K-means clustering algorithm is used to cluster each superpixel to form different target regions; Step D: Based on the results of step C, select items with an aspect ratio greater than the threshold. Long, strip-shaped targets that satisfy the condition that the region size exceeds a threshold. The length direction is calculated, and the average motion vector of the center of the selected elongated target, which is the blade, is calculated based on the optical flow analysis results. Step E: When the similarity along the length direction of the elongated target is greater than the threshold If the similarity of the average motion vector direction at the center of all elongated targets is greater than a threshold, it is judged as a vertical axis fan; if the length direction of the elongated target has a certain angle and the average motion vector direction at the center of the elongated target has a certain angle with each other, it is judged as a horizontal axis fan; other states cannot be distinguished. Step F: If the current analysis fails to identify the type, repeat steps B-E until identification is successful. Once successful, stop the above type identification process and begin the motion measurement phase.

Images