A dual-field-of-view, wide-area, high-precision particle size measurement device and method for high-speed cloud and fog fields

Abstract

The present invention discloses a high-precision device and method for measuring particle size in a wide range with dual fields of view for high-speed cloud and fog fields, belonging to the field of measuring liquid water content in cloud and fog flow fields. Based on an optical magnification device, the present invention synchronously acquires visualized images of droplets in the same field of view at different magnifications. It utilizes a ResNet neural network-GAN network dual-network coupled recognition to achieve high-precision measurement of droplet size in high-speed cloud and fog fields, solving the problems of poor time responsiveness, low measurement accuracy, and low reliability of backtracking algorithms associated with traditional wind tunnel cloud and fog field parameter measurement devices. Furthermore, the optical magnification device is used to capture optical images, which can simultaneously meet the testing requirements for droplet size and a large field of view. Furthermore, the ResNet neural network-GAN network dual-network coupled recognition is used to learn calibration data and construct correlations between droplet images at different magnifications, fully extracting potential features from the image and improving test accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of liquid water content measurement in cloud and fog flow fields, and in particular to a device and method for high-precision measurement of particle size with a dual field of view and a wide range for high-speed cloud and fog fields. Background Art

[0002] The precise identification and measurement of liquid droplets in a two-phase flow field plays a key role in the study of basic combustion issues, flow field characteristics analysis, and aircraft engine R&D and design. Among them, the simulation and measurement of aircraft engine intake conditions still face many technical difficulties that need to be overcome. The national military standards and airworthiness regulations have strict requirements on the control accuracy of intake simulation parameters, especially the measurement accuracy of parameters such as liquid water content (LWC) and water droplet size (MVD) in the intake cloud flow field. However, for the intake environment simulation of high-altitude platforms, there is still a lack of reliable cloud parameter measurement technology, which restricts my country's aircraft engine intake test assessment and airworthiness verification capabilities.

[0003] Currently, there are many methods for measuring liquid water content (LWC), including the ice skate method, rotating multi-cylinder measurement method, and hot wire measurement method. The ice skate method is a commonly used LWC calibration method both domestically and internationally, but its overall efficiency is relatively low, and it is mainly used for mutual comparison and verification of calibration results. The rotating multi-cylinder measuring instrument is the simplest and most reliable instrument for measuring water droplet parameters. Its use has certain limitations: for example, when the surface temperature of the cylinder approaches or exceeds 0°C, it cannot be measured. The hot wire measurement method is mature and widely used. It calculates the liquid water content by evaporating water droplets on the surface of a resistance wire, which causes the temperature and impedance of the resistance wire to change. Its measurement range is usually 10-40μm.

[0004] There are four main methods used abroad to measure the mean volumetric particle size (MVD) of droplets in wind tunnel flow fields: forward scattering spectrometer (FSSP), optical array meter (OAP), phase Doppler particle analyzer (PDPA), and Malvern particle size analyzer (MPSA). The forward scattering spectrometer (FSSP) is widely used for ground test equipment and flight test cloud measurement. If there are too many small droplets in its probe volume, it will cause counting errors and particle size measurement errors. The optical array meter (OAP) is a commonly used instrument for measuring the diameter of cloud droplets. It has a wide range and many models. It is often used to measure small droplets with a diameter greater than 100μm. The phase Doppler particle analyzer (PDPA) is manufactured by Aerodynamic Measurement Co., Ltd. in the United States. It is used for measuring the particle size of small droplets in wind tunnels and flight tests. It is not suitable for measuring spray fields with high number density. The system has high requirements for the test environment and the optical path adjustment is cumbersome. The Malvern Particle Sizer (MPSA) is widely used to analyze the particle size of fuel atomized droplets. The test results in the spray area with an excessively large density field are subject to large deviations. Multiple refraction, deflection, and halation of light can affect the test results.

[0005] Due to the late start of domestic aircraft research, most of it has focused on theoretical research and numerical simulation. To date, there are only a few wind tunnels in China. However, with the development of independent aircraft design in recent years, the demand for aircraft anti-icing design has grown rapidly. Regarding cloud field parameter testing and research, specialized optical testing instruments from abroad are expensive, difficult to calibrate, and have low measurement efficiency. China lacks independently developed cloud field parameter measurement methods, instrumentation, and measurement software for the rapid and accurate measurement of LWC and MVD in high-speed cloud fields. Therefore, there is an urgent need to develop synchronous, rapid, and accurate measurement technology for LWC and MVD in high-speed cloud fields to support the development of simulation technology for aircraft engine inlet icing environments in my country. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a high-precision measurement device and method for particle size in a wide range with dual fields of view for high-speed cloud and fog fields.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A high-precision dual-field-of-view wide-range particle size measurement device for high-speed cloud and fog fields, comprising a dual-field-of-view wide-range particle size recognition device;

[0009] The dual-field-of-view wide-area particle size high-precision measurement device includes a laser beam expander, an optical magnifying device and a computer; the optical magnifying device includes a beam splitter, a primary magnifying glass is connected to the light inlet of the beam splitter, a primary magnifying glass magnification adjustment knob is provided on the side of the primary magnifying glass, one optical path outlet of the beam splitter is connected to a first CCD camera through a secondary magnifying glass at a small magnification end, and the other optical path outlet is connected to a second CCD camera through a secondary magnifying glass at a large magnification end, the first CCD camera and the second CCD camera are both provided with a signal transmission interface and a power cord, the signal transmission interface is connected to a computer, and an identification module is provided in the computer; the laser beam expander inlet is used to receive laser light, the laser beam expander and the primary magnifying glass are arranged relative to each other, the optical axes of the laser beam expander and the primary magnifying glass coincide in the same horizontal plane, and the distance between the two is 50 to 80 cm;

[0010] When identification is performed, the droplets in the cloud field to be detected enter the detection area between the laser beam expander and the primary magnifying glass;

[0011] The recognition module is used to receive the acquisition signals of the first CCD camera and the second CCD camera, that is, two groups of images of the same area at different magnifications, and based on the two groups of images, segment and extract the image of the detection area overlapping with the large magnification image in the small magnification image as the small magnification processed image; perform image registration on the small magnification processed image and the large magnification image, that is, obtain the droplets identified at the same position in the two images at the same time, extract the droplets in different images respectively and store them in a fixed size, and store them in the small magnification droplet image group and the large magnification droplet image group respectively; based on the conditional GAN ​​network, use the small magnification droplet image group as input to generate a realistic magnified image, use the large magnification droplet image group as the identification object, judge whether the generated realistic magnified image is consistent with the large magnification droplet image group, and continuously train until an association network model between the different magnification droplet images is established, so that the realistic magnified image is consistent with the large magnification droplet image group;

[0012] All droplets in the image of the small magnification array are segmented, extracted and saved, and input into the established association network model of the different magnification droplet images to generate realistic magnified droplet images;

[0013] Construct a ResNet neural network, input the calibration images in the calibration database into the ResNet neural network, and establish an accurate recognition network for droplet particle size;

[0014] Inputting the realistic magnified droplet image into the precise recognition network to obtain particle size information of each droplet in the image group with a small magnification;

[0015] The cloud and fog field incoming flow parameters LWC and MVD are calculated based on the particle size information.

[0016] Furthermore, the laser at the entrance of the laser beam expander is obtained by two-stage reflection of an Nd:YAG laser.

[0017] A high-precision method for measuring particle size in a wide range with dual fields of view for high-speed cloud and fog fields comprises the following steps:

[0018] (1) Install an etched circular calibration plate between the laser beam expander and the primary magnifying mirror. The plane of the etched circular calibration plate is parallel to the laser outlet section of the laser beam expander. Adjust the secondary magnifying mirror at the small magnification end of the optical magnifying device and the secondary magnifying mirror at the large magnification end of the optical magnifying device so that their test areas are respectively in the preset areas. Fix the magnification parameters of the secondary magnifying mirror at the small magnification end and the secondary magnifying mirror at the large magnification end, and perform static calibration.

[0019] The etched circular calibration plate is moved by a three-dimensional micro-displacement platform to obtain a calibration database of particles of different sizes in three different states: in-focus, positive defocus, and negative defocus. Images are captured by a first CCD camera and a second CCD camera and stored in a computer.

[0020] (2) Dismantle the etched circular calibration plate and the three-dimensional micro-displacement stepping platform, and move the above-mentioned dual-field-of-view wide-range particle size high-precision measurement device for high-speed cloud and fog fields so that the droplets in the cloud and fog fields to be measured are distributed in the detection area between the laser beam expander and the primary magnifying glass;

[0021] (3) using the second CCD camera and the first CCD camera to collect image data at the small magnification end and the large magnification end, respectively, and storing the data in groups in a computer;

[0022] (4) acquiring the acquisition signals of the first CCD camera and the second CCD camera, i.e., two sets of images of the same area at different magnifications; based on the two sets of images, segmenting and extracting the image of the detection area in the small magnification image that overlaps with the large magnification image, and using it as the small magnification processed image; performing image registration on the small magnification processed image and the large magnification image, i.e., obtaining the droplets identified at the same position in the two images at the same moment, extracting the droplets in different images respectively and storing them in a fixed size, respectively, and storing them in the small magnification droplet image group and the large magnification droplet image group; based on the conditional GAN ​​network, taking the small magnification droplet image group as input, generating a realistic magnified image, taking the large magnification droplet image group as the identification object, judging whether the generated realistic magnified image is consistent with the large magnification droplet image group, and finally establishing an association network model between the different magnification droplet images, so that the realistic magnified image is consistent with the large magnification droplet image group;

[0023] All droplets in the image of the small magnification array are segmented, extracted and saved, and then imported into the established association network model of the different magnification droplet images to generate realistic magnified droplet images;

[0024] Construct a ResNet neural network, import calibration images from the calibration database into the network, and establish an accurate recognition network for droplet particle size;

[0025] Importing the realistic magnified droplet image into the precise recognition network to obtain particle size information of each droplet in the image of the small magnification image group;

[0026] The cloud and fog field incoming flow parameters LWC and MVD are calculated based on the particle size information.

[0027] Furthermore, in step (4), the particle size information is used to calculate the cloud and fog field incoming flow parameter LWC as follows:

[0028]

[0029] Where V CSA The camera capture area, is the volume of the i-th identified droplet in the shooting area;

[0030] V CSA =S CSA ·D max (2)

[0031] Where D max is the maximum droplet size in the field of view, S CSA is the field of view area perceived by the camera pixels and magnification;

[0032]

[0033] Where D i is the particle size of the i-th identified droplet;

[0034] Furthermore, in step (4), the particle size information is used to calculate the cloud field flow parameter MVD, specifically:

[0035]

[0036] Where n is the total number of droplets in the field of view.

[0037] Furthermore, in step (1), the test area at the small magnification end reaches 1mm 3 above.

[0038] Furthermore, in step (1), the resolution of the tested particle size at the large magnification end reaches below 1 μm.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention provides a dual-field-of-view wide-area particle size high-precision measurement device and method for high-speed cloud and fog fields. Based on an optical magnification device, visualization images of droplets in the same field of view with different magnifications are obtained synchronously. A ResNet neural network-GAN network dual-network coupling recognition algorithm is used to achieve high-precision measurement of droplet size in high-speed cloud and fog fields, thereby solving the problems of poor time response, low measurement accuracy, low reliability of backtracking algorithm, and inability to measure quickly in traditional wind tunnel cloud and fog field parameter measurement devices. At the same time, the optical magnification device is used to capture optical images, which can simultaneously meet the test requirements of droplet size and large field of view, solving the problems of traditional shadow method technology such as wide droplet size range, large measurement field of view, and lack of space. The difficult problem of balancing high resolution and low resolution is solved; the use of ResNet neural network-GAN network dual network coupling recognition for calibration data learning and the construction of correlation relationships between droplet images of different magnifications can utilize large-scale data to fully extract potential features in the image, avoid excessive reliance on and limitations of parameters such as grayscale in traditional image measurement technology, and thus improve test accuracy; the present invention adopts visual measurement technology, the measurement results are highly accurate, and the measurement process has a clear physical basis; the device proposed in the present invention has a simple structure, modular design, and is portable. It is suitable for airflow conditions with different air velocities, and high-precision and rapid measurement of LWC and MVD in a wide particle size range and a large field of view in cloud and fog fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a structural diagram of the dual-field-of-view wide-area particle size identification device of the present invention;

[0042] Figure 2 This is a structural diagram of the cloud field droplet generation module of the present invention;

[0043] Figure 3 This is a specific embodiment diagram of the cloud field droplet LWC / MVD measurement of the present invention;

[0044] Figure 4 This is a structural diagram of an optical magnification device;

[0045] Figure 5 This is a flowchart of the ResNet neural network-GAN network dual network coupling identification of the present invention.

[0046] Among them: 01-first reflector support seat; 02-Nd:YAG laser; 03-3D micro-displacement stepper; 04-etched circular calibration plate; 05-beam splitter; 06-secondary magnifying mirror 1 at the small magnification end; 07-first CCD camera; 08-first reflector; 09-second reflector support seat; 10-second reflector; 11-laser beam expander; 12-primary magnifying mirror; 13-secondary magnifying mirror at the large magnification end; 14-second CCD camera; 15-cooling section insulation layer; 16-Cooling section; 17-Water supply tank; 18-Gradually converging steady flow section; 19-First nozzle; 20-Air duct; 21-Blower; 22-Second nozzle; 23-Blower air inlet; 24-Water supply pipeline; 25-First compressed air delivery pipe; 26-Blower air outlet; 27-Second compressed air delivery pipe; 28-Gradually diverging nozzle; 29-Droplets; 30-Computer; 31-Primary magnifying glass magnification adjustment knob; 32-Second magnifying glass magnification adjustment knob; 33-Signal transmission interface; 34-Power cord. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0048] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] A high-precision particle size measurement device with a dual field of view and wide range for high-speed cloud and fog fields, comprising an optical magnification and signal acquisition module, a high-energy laser light source module, a cloud and fog field droplet generation module, a calibration device, and a particle size precision identification and analysis module. The present invention generates a cloud and fog field environment with parameters such as LWC / MVD to be measured through the cloud and fog field droplet generation module. Based on the ultra-bright low-frequency light source provided by the high-energy laser light source module, the optical magnification and signal acquisition module photographs the cloud and fog field environment, and transmits the resulting image to the particle size precision identification and analysis module. The ResNet neural network-GAN network dual-network coupling recognition algorithm within the module is used to accurately identify droplet particle size and measure parameters such as LWC / MVD. The calibration device consists of an etched circular calibration plate and a three-dimensional micro-displacement platform. The particle size distribution of the circular droplets in the calibration plate is 0.5μm-3000μm, and the displacement step length of the three-dimensional micro-displacement platform in the xyz directions is 1μm. The wind tunnel cloud field droplet generation module, consisting of a blower, nozzle, air duct, and cooling section, is a wind tunnel cloud field simulation device capable of generating tiny droplets in high-speed horizontal motion. The high-energy laser light source module, consisting of a short-pulse, low-frequency laser and a laser beam expander, provides a sufficiently bright light source for the optical magnification and signal acquisition module (the unit illumination of conventional continuous illumination sources reaches 200,000 LUX, but this still does not meet testing requirements). The beam expander also ensures that the light source energy does not damage the optical components. The optical magnification and signal acquisition module consists of a magnifying lens and two CCD cameras. After the cloud field passes through the magnifying lens and two CCD cameras, two images with different magnifications are generated. These two images capture droplet information within the same area of ​​the cloud field. This image information is transmitted to the particle size precision identification and analysis module for precise droplet size identification and measurement of parameters such as LWC / MVD. During measurement, the high-energy laser light source module and the optical amplification and signal acquisition module are placed at either end of the wind tunnel's cloud field droplet generation module, approximately 10-100 cm from its exit cross-section. The central optical axes of the two modules are aligned in the same horizontal plane. Finally, a communication fiber optic cable connects the optical amplification and signal acquisition module to the particle size precision identification and analysis module.

[0050] The main body of the particle size precision identification and analysis module is a high-performance computer. The computer performs LWC parameter measurement based on the ResNet neural network-GAN network dual-network coupling identification algorithm. The detailed process is as follows: First, after long-term shooting through the test system (composed of an optical magnification and signal acquisition module, a high-energy laser light source module, and a wind tunnel cloud field droplet generation module), two sets of experimental images of the same area at different magnifications are obtained. Among them, the experimental images in the small magnification array have a small magnification and a larger shooting area; the experimental images in the large magnification array have a large magnification and a smaller shooting area. Therefore, the actual test area of ​​the large magnification image is only the central area of ​​the small magnification image. In the second step, the algorithm segments and extracts the test area in the small magnification image that overlaps with the large magnification image, and stores it as a small magnification processed image; in the third step, the algorithm performs image registration on the small magnification processed image and the large magnification image, that is, there should be identifiable droplets at the same position in the two images at the same time. The above droplets are extracted and stored in a fixed size, respectively, in the small magnification droplet image group and the large magnification droplet image group; in the fourth step, based on the conditional GAN ​​network, the algorithm uses the small magnification droplet image group as output to generate a realistic magnified image, and uses the large magnification droplet image group as the identification object to determine whether the generated realistic magnified image is consistent with the large magnification droplet image group, and finally establishes a different magnification droplet image. The associated network model between them makes the realistic magnified image consistent with the large magnification droplet image group; the fifth step is to segment, extract and save all the droplets in the experimental image of the small magnification group, and then import them into the associated network model of the different magnification droplet image established in the fourth step to generate a realistic magnified droplet image; the sixth step is to use the static calibration plate and the large magnification in the first step to establish a calibration database for the droplets; the seventh step is to construct a ResNet neural network, import the calibration image in the calibration database into the network, and establish an accurate recognition network for droplet particle size; the eighth step is to import the realistic magnified droplet image in the fifth step into the accurate recognition network established in the seventh step to obtain the particle size information of each droplet in the experimental image in the small magnification image group, and then calculate parameters such as LWC.

[0051] This precise particle size identification algorithm based on optical magnification can, on the one hand, solve the physical problem of a small measurement area when the magnification is too large; on the other hand, since droplets have simple geometric shapes such as circles or ellipses, their features can be more easily extracted and learned in neural networks. Therefore, constructing a precise particle size identification algorithm can fully utilize the deep learning capabilities of neural networks.

[0052] The optical magnification and signal acquisition module is a key component of the dual-field-of-view wide-area particle size identification device. Achieving a magnification of the same area can be achieved through the following different technologies. In the first technical solution, the amplification is achieved by a magnifying glass with an adjustable optical magnification, and the same field of view is photographed by adjusting the magnification. This solution has the advantages of a simple optical magnification device, a simple internal optical system, and good imaging quality. However, since the optical magnification needs to be adjusted manually, the field of view cannot be optically amplified at the same time, so it is only suitable for photographing static or steady-state objects. In the second technical solution, the amplification is achieved by a basic magnifying glass combined with a beam splitter, which initially amplifies the image in the field of view and divides it into two. Then, secondary magnifying glasses are set at the two exit sections of the beam splitter. By adjusting the magnification of the secondary magnifying glasses, the synchronous amplification function of the same field of view is achieved. The optical magnification device in this solution is complex, with multiple internal optical systems that require adjustment and matching, and the imaging quality is significantly lower than that of the first technical solution. However, because it can achieve synchronous magnification within the same field of view, it can perform high-precision measurements of transient physical objects such as high-speed non-steady states and turbulent fields. This present invention adopts the second technical solution.

[0053] With respect to the high-speed cloud and fog field dual-field-of-view wide-range particle size high-precision measurement device proposed in the present invention, the present invention proposes a method for measuring the cloud and fog field incoming flow parameters LWC / MVD using the above device, the steps of which are as follows:

[0054] The main steps for measuring cloud field flow parameters such as LWC / MVD include: laboratory bench setup, static calibration, cloud simulation, image acquisition, accurate particle size identification, and LWC / MVD measurement. Specifically, the following steps are involved:

[0055] A. Experimental bench setup: According to the test requirements of the backlight method, the high-energy laser light source module and the optical magnification and signal acquisition module are placed in a facing installation. In order to ensure that the laser does not reflect into the laser resonant cavity and cause damage to the laser, the short-pulse-width low-frequency laser should be corrected by a reflector after emission and introduced into the laser beam expander. The laser outlet cross-section of the laser beam expander should be installed parallel to the entrance cross-section of the optical magnification device, leaving a remaining space of 50cm-80cm in the middle for installing devices such as the calibration device and the cloud field droplet generation module. In addition, it is necessary to ensure that the optical axes of the laser beam expander and the optical magnification device coincide in the same horizontal plane;

[0056] B. Static calibration: Install the calibration device between the laser beam expander and the optical magnification device. The plane of the etched circular calibration plate is parallel to the laser beam expander laser outlet section. By adjusting the secondary magnifying mirror of the optical magnification device, ensure that the test area at the small magnification end reaches 1mm. 3The above measures ensure that the particle size resolution at the maximum magnification end is below 1μm. The magnification parameters of the optical magnification device are fixed and static calibration is performed. Static calibration uses a three-dimensional micro-displacement platform to control the movement of the etched calibration plate to obtain a calibration database of particles of different particle sizes in three different states: in-focus, positive defocus, and negative defocus. Images are collected and stored using two CCD cameras.

[0057] C. Cloud simulation: After disassembling the calibration device, install the wind tunnel cloud field droplet generation module. The center cross-section of the cooling section outlet of the module should coincide with the plane of the calibration plate. The specific installation position of the module should be determined based on the actual test area. Open the nozzle and cooling section. After droplets are generated, turn on the blower to ensure stable droplet generation within the test area.

[0058] D. Image acquisition: To maintain the normal working state of the cloud field droplet generation module, two CCD cameras are used to collect image data at the small magnification end and the large magnification end respectively, and save them in groups;

[0059] E. Accurately identify the particle size and store the two sets of experimental images of the same area at different magnifications as a small magnification array and a large magnification array respectively. The test area overlapping with the large magnification image in the small magnification image is segmented and extracted, and stored as a small magnification processing image; the small magnification processing image and the large magnification image are image-aligned, and the same droplets in a pair of images are identified, extracted and stored with a fixed size, and stored in the small magnification droplet image group and the large magnification droplet image group respectively; based on the conditional GAN ​​network, the small magnification droplet image group is used as input to establish an association network model between the different magnification droplet images, so that the realistic magnified image is consistent with the large magnification droplet image group; all droplets in the experimental image in the small magnification group are segmented, extracted and saved, and then imported into the association network model of the different magnification droplet image to generate a realistic magnified droplet image; the calibration image in the calibration database is imported into the ResNet neural network to establish a precise recognition network for droplet particle size; the realistic magnified droplet image is imported into the precise recognition network for droplet particle size to obtain the particle size information D of each droplet in the experimental image in the small magnification image group i ;

[0060] F. For LWC / MVD measurement, the droplet size statistical information obtained based on optical magnification and the ResNet neural network-GAN network dual network coupling recognition algorithm is processed secondary according to Equations 1, 2, 3, and 4 to complete the LWC and MVD measurement results of the cloud field, and then output and save them.

[0061]

[0062] Where VCSA is the camera shooting area, which can be calculated by formula 2 in this method. is the volume of each identified droplet in the shooting area. According to the spherical droplet assumption, its volume is calculated by Equation 3.

[0063] V CSA =S CSA ·D max (2)

[0064] Where D max is the maximum droplet size in the field of view, S CSA It is the field of view area perceived by the camera pixels and magnification.

[0065]

[0066] Where D i is the particle size of the i-th identified droplet.

[0067]

[0068] Where n is the total number of droplets in the field of view.

[0069] The present invention is described in further detail below with reference to the accompanying drawings:

[0070] See also Figure 1 , Figure 1This is a structural diagram of the dual-field-of-view wide-range particle size identification device of the present invention, which includes a first reflector support seat 01, an Nd:YAG laser 02, a three-dimensional micro-displacement stepping stage 03, an etched circular calibration plate 04, a beam splitter 05, a secondary magnifying mirror at a small magnification end 06, a first CCD camera 07, a first reflector 08, a second reflector support seat 09, a second reflector 10, a laser beam expander 11, a primary magnifying mirror 12, a secondary magnifying mirror at a large magnification end 13 and a second CCD camera 14; wherein, the light path of the Nd:YAG laser 02 enters the laser beam expander 11 after being reflected by the first reflector 08 and the second reflector 10; the laser beam expander 11 and the primary magnifying mirror 12 are arranged opposite to each other, the optical axes of the laser beam expander 11 and the primary magnifying mirror 12 coincide in the same horizontal plane, and the distance between the two is 50 to 80 cm. A beam splitter 05 is provided on the light beam exit side of the primary magnifying mirror 12, and the beam splitter 05 divides the incoming light path into two, one of which is connected to the secondary magnifying mirror 06 at the small magnification end of the optical different magnification amplification device, and the secondary magnifying mirror 06 at the small magnification end is connected to the first CCD camera 07, and the other is connected to the secondary magnifying mirror 13 at the large magnification end of the optical different magnification amplification device, and the secondary magnifying mirror 13 at the large magnification end is connected to the second CCD camera 14, and the data of the first CCD camera 07 and the second CCD camera 14 are transmitted to the computer 30;

[0071] An etched circular calibration plate 04 is provided between the laser beam expander 11 and the primary magnifying mirror 12 , and the etched circular calibration plate 04 is connected to a three-dimensional micro-displacement stepping stage 03 .

[0072] See also Figure 2 , Figure 2 The structure diagram of the cloud field droplet generation module of the present invention; the cloud field droplet generation module of the present invention includes a cooling section insulation layer 15, a cooling section 16, a water supply tank 17, a gradually contracting steady flow section 18, a first nozzle 19, an air duct 20, a blower 21, a second nozzle 22, a blower air inlet 23, a water supply pipeline 24, a first compressed air delivery pipe 25, a blower air outlet 26, a second compressed air delivery pipe 27 and a gradually diverging nozzle 28; the blower 21 is provided with a blower air inlet 23, and the blower air outlet 26 of the blower 21 is connected to the first A compressed air delivery pipe 25, a gradually diverging nozzle 28 is connected between the first compressed air delivery pipe 25 and the second compressed air delivery pipe 27, an air duct 20 is provided at the end of the second compressed air delivery pipe 27, and a first nozzle 19 and a second nozzle 22 are provided on both side walls of the air duct 20. The first nozzle 19 and the second nozzle 22 are connected to the water supply tank 17 through a water supply pipe 24, and the outlet of the air duct 20 is connected to the gradually convergent flow stabilizing section 18 and the cooling section 16 in sequence. A cooling section insulation layer 15 is provided on the outside of the cooling section 16, and droplets 29 are generated at the outlet of the cooling section 16.

[0073] See also Figure 3 , Figure 3 This is a specific implementation diagram of the cloud field droplet LWC / MVD measurement of the present invention. The outlet of the cooling section 16 of the wind tunnel cloud field droplet generation module is within the effective detection area between the laser beam expander 11 and the primary magnifying mirror 12.

[0074] See also Figure 4 , Figure 4 This is a structural diagram of an optical different-magnification magnification device. It can be seen that the optical different-magnification magnification device includes a beam splitter 05. A primary magnifying glass 12 is connected to the light inlet of the beam splitter 05. A primary magnifying glass magnification adjustment knob 31 is provided on the side of the primary magnifying glass 12. One optical path outlet of the beam splitter 05 is connected to a first CCD camera 07 through a secondary magnifying glass 06 at a small magnification end, and the other optical path outlet is connected to a second CCD camera 14 through a secondary magnifying glass 13 at a large magnification end. Both the first CCD camera 07 and the second CCD camera 14 are provided with a signal transmission interface 33 and a power line 34.

[0075] The invention provides a dual-field-of-view wide-range particle size identification device in a high-speed cloud and fog field, comprising an Nd:YAG laser 2, a laser beam expander 11, a primary magnifying glass 12, a secondary magnifying glass 6 at a small magnification end, a secondary magnifying glass 13 at a large magnification end, a first CCD camera 7 and a second CCD camera 14.

[0076] The droplets 29 are generated by the tiny droplets generated by the first nozzle 19 and the second nozzle 22 through the cooling section 16. The airflow provided by the blower 21 passes through the blower outlet 26, the compressed air delivery pipe 25, the gradually diverging nozzle 28, the compressed air delivery pipe 27, the air duct 20, the gradually converging steady flow section 18 and the cooling section 16, and blows the droplets 29 into the effective detection area between the laser beam expander 11 and the primary magnifying glass 12.

[0077] After the droplet information is captured by the primary magnifying glass, the beam splitter 5 at its rear end divides the signal into two parts. The first CCD camera 7 and the second CCD camera 14 are respectively installed behind the secondary magnifying glass 6 at the small magnification end and the secondary magnifying glass 13 at the large magnification end, and are used to capture the droplet images with different magnifications. The obtained images are imported into the computer 30 through the signal transmission interface 33, and the ResNet neural network-GAN network dual network coupling recognition algorithm is used to accurately calculate the droplets.

[0078] Specifically, such as Figure 1 As shown in FIG, the dual-field-of-view wide-range particle size identification device proposed in the present invention for high-speed cloud and fog fields, taking droplets generated in a wind tunnel simulation device in a laboratory as an example, when the device is used, LWC and MVD measurements include the following specific steps:

[0079] (1) According to the test requirements of the backlight method, the laser beam expander 11 in the high-energy laser light source module and the primary magnifying mirror 12 in the optical magnification and signal acquisition module are placed in a facing installation manner. In order to ensure that the laser does not reflect into the resonant cavity of the laser 2 and cause damage to the laser 2, the short-pulse-width low-frequency laser should be corrected by the reflector 8 and the reflector 9 after emission and introduced into the laser beam expander 11. The laser outlet cross section of the laser beam expander 11 should be installed parallel to the entrance cross section of the primary magnifying mirror 12, leaving a distance of 50 to 80 cm in between for the installation of devices such as the etched circular calibration plate 4 and the cloud field droplet generation module. In addition, it is necessary to ensure that the optical axes of the laser beam expander 11 and the primary magnifying mirror 12 coincide in the same horizontal plane;

[0080] (2) Install the etched circular calibration plate 4 between the laser beam expander 11 and the primary magnifying mirror 12. The plane of the etched circular calibration plate 4 is parallel to the laser exit section of the laser beam expander 11. Adjust the secondary magnifying mirror 6 at the small magnification end of the optical magnification device to ensure that the test area at the small magnification end reaches 1mm. 3 The above steps adjust the secondary magnifying mirror 13 at the large magnification end of the optical magnification device to ensure that the test particle size resolution at the large magnification end is less than 1 μm. The magnification parameters of the secondary magnifying mirror 6 at the small magnification end and the secondary magnifying mirror 13 at the large magnification end are fixed, and static calibration is performed. The static calibration controls the movement of the etched circular calibration plate 4 by the three-dimensional micro-displacement platform 3 to obtain a calibration database of particles of different particle sizes (0.5 μm-200 μm) in three different states: in-focus, positive defocus, and negative defocus. The images are captured by the first CCD camera 7 and the second CCD camera 14 and stored in the computer 30.

[0081] (3) After removing the etched circular calibration plate 4 and the three-dimensional micro-displacement stepping platform 3, install the cloud field droplet generation module. The center cross-section of the cooling section 16 outlet of the module coincides with the plane of the etched circular calibration plate 4. The specific installation position of the module is determined according to the actual area to be measured. Open the first nozzle 19, the second nozzle 22 and the cooling section 16. After the droplets 29 are generated, turn on the blower 21 to ensure that the droplets 29 can be stably generated in the area to be measured.

[0082] (4) Maintaining the working state of the cloud field droplet generation module, using the second CCD camera 14 and the first CCD camera 7 to collect image data at the small magnification end and the large magnification end, respectively, and saving the data in groups in the computer 30;

[0083] (5) The two sets of experimental images acquired at different magnifications of the same area are stored as a small magnification array and a large magnification array respectively. The test area overlapping with the large magnification image in the small magnification image is segmented and extracted, and stored as a small magnification processing image; the small magnification processing image and the large magnification image are image-aligned, and the same droplets in a pair of images are identified, extracted and stored with a fixed size, and stored in the small magnification droplet image group and the large magnification droplet image group respectively; based on the conditional GAN ​​network, the small magnification droplet image group is used as input to establish an association network model between the different magnification droplet images, so that the realistic magnified image is consistent with the large magnification droplet image group; all droplets in the experimental image in the small magnification group are segmented, extracted and saved, and then imported into the association network model of the different magnification droplet image to generate a realistic magnified droplet image; the calibration image in the calibration database is imported into the ResNet neural network to establish a precise recognition network for droplet particle size; the realistic magnified droplet image is imported into the precise recognition network for droplet particle size to obtain the particle size information D of each droplet in the experimental image in the small magnification image group i ;

[0084] (6) The droplet size statistical information obtained based on the optical magnification and ResNet neural network-GAN network dual network coupling recognition algorithm is processed secondary according to Equations 1, 2, 3, and 4 to complete the LWC measurement results of the cloud field, and output and save them;

[0085]

[0086] Where V CSA is the camera shooting area, which is calculated using Formula 2 in the present invention. is the volume of the i-th identified droplet in the shooting area. According to the spherical droplet assumption, its volume is calculated by Equation 3;

[0087] V CSA =S CSA ·D max (2)

[0088] Where D max is the maximum droplet size in the field of view, S CSA It is the field of view area perceived by the camera pixels and magnification.

[0089]

[0090] Where D i is the particle size of the i-th identified droplet;

[0091]

[0092] Where n is the total number of droplets in the field of view.

[0093] (7) Turn off the blower 21, the second CCD camera 14, the first CCD camera 7 and the Nd:YAG laser 2, disassemble the optical components such as the reflector 8, the reflector 9, the laser beam expander 11, the primary magnifying glass 12, the beam splitter 5, the secondary magnifying glass 6 at the small magnification end and the secondary magnifying glass 13 at the large magnification end, place them in a dustproof room, and clean the laboratory environment.

[0094] The above content is only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A high-precision dual-field-of-view wide-range particle size measurement device for high-speed cloud and fog fields, characterized in that: Including dual-field wide-area particle size identification device; The dual-field wide-area particle size high-precision measuring device comprises a laser beam expander (11), an optical magnifying device and a computer (30); the optical magnifying device comprises a beam splitter (05), a primary magnifying glass (12) is connected to the light inlet of the beam splitter (05), a primary magnifying glass magnification knob (31) is provided on the side of the primary magnifying glass (12), one optical path outlet of the beam splitter (05) is connected to a first CCD camera (07) via a secondary magnifying glass (06) at a small magnification end, and the other optical path outlet is connected to a second CCD camera (07) via a secondary magnifying glass (13) at a large magnification end. There is a second CCD camera (14), the first CCD camera (07) and the second CCD camera (14) are both provided with a signal transmission interface (33) and a power line (34), the signal transmission interface (33) is connected to a computer (30), and the computer (30) is provided with an identification module; the laser beam expander (11) entrance is used to receive laser light, the laser beam expander (11) and the primary magnifying mirror (12) are arranged relative to each other, the optical axes of the laser beam expander (11) and the primary magnifying mirror (12) coincide in the same horizontal plane, and the distance between the two is 50 to 80 cm; When identification is performed, the cloud field droplets to be detected enter the detection area between the laser beam expander (11) and the primary magnifying glass (12); The recognition module is used to receive the acquisition signals of the first CCD camera (07) and the second CCD camera (14), that is, two groups of images of the same area at different magnifications, and based on the two groups of images, segment and extract the image of the detection area that overlaps with the large magnification image in the small magnification image as the small magnification processed image; perform image registration on the small magnification processed image and the large magnification image, that is, obtain the droplets identified at the same position in the two images at the same time, extract the droplets in different images respectively and store them in a fixed size, and store them in the small magnification droplet image group and the large magnification droplet image group respectively; Based on the conditional GAN ​​network, a group of small magnification droplet images is used as input to generate realistic magnified images. A group of large magnification droplet images is used as the identification object to determine whether the generated realistic magnified images are consistent with the large magnification droplet image group. Training is continued until an association network model between the different magnification droplet images is established, so that the realistic magnified images are consistent with the large magnification droplet image group. All droplets in the image of the small magnification array are segmented, extracted and saved, and input into the established association network model of the different magnification droplet images to generate realistic magnified droplet images; Construct a ResNet neural network, input the calibration images in the calibration database into the ResNet neural network, and establish an accurate recognition network for droplet particle size; Inputting the realistic magnified droplet image into the precise recognition network to obtain particle size information of each droplet in the image group with a small magnification; The cloud and fog field incoming flow parameters LWC and MVD are calculated based on the particle size information.

2. The high-precision dual-field-of-view wide-range particle size measurement device for high-speed cloud and fog fields according to claim 1 is characterized in that: The laser light at the entrance of the laser beam expander (11) is obtained by a Nd:YAG laser (02) through two-stage reflection.

3. A high-precision dual-field-of-view wide-area particle size measurement method for high-speed cloud and fog fields, characterized in that: A high-precision dual-field-of-view wide-area particle size measurement device for high-speed cloud and fog fields according to claim 1 or 2; The method comprises the following steps: (1) Installing the etched circular calibration plate (4) between the laser beam expander (11) and the primary magnifying mirror (12), with the plane of the etched circular calibration plate (4) being parallel to the laser exit section of the laser beam expander (11), adjusting the secondary magnifying mirror (6) at the small magnification end of the optical magnifying device and the secondary magnifying mirror (13) at the large magnification end of the optical magnifying device so that the test areas of the two are respectively in the preset areas, fixing the magnification parameters of the secondary magnifying mirror (6) at the small magnification end and the secondary magnifying mirror (13) at the large magnification end, and performing static calibration; The movement of the etched circular calibration plate (4) is controlled by a three-dimensional micro-displacement platform (3) to obtain a calibration database of particles of different particle sizes in three different states: in-focus, positive defocus, and negative defocus. Images are collected by a first CCD camera (7) and a second CCD camera (14) and stored in a computer (30); (2) dismantling the etched circular calibration plate (4) and the three-dimensional micro-displacement stepping platform (3), and moving the dual-field-of-view wide-area particle size high-precision measuring device for high-speed cloud and fog fields as described in claim 1 or 2, so that the droplets in the cloud and fog field to be measured are distributed in the detection area between the laser beam expander (11) and the primary magnifying glass (12); (3) using the second CCD camera (14) and the first CCD camera (7) to respectively collect image data at the small magnification end and the large magnification end, and storing the data in groups in a computer (30); (4) acquiring the acquisition signals of the first CCD camera (07) and the second CCD camera (14), i.e., two sets of images of the same area at different magnifications; based on the two sets of images, segmenting and extracting the image of the detection area in the small magnification image that overlaps with the large magnification image, and using it as the small magnification processed image; performing image registration on the small magnification processed image and the large magnification image, i.e., acquiring the droplets identified at the same position in the two images at the same moment, extracting the droplets in different images respectively and storing them in a fixed size, and storing them in the small magnification droplet image group and the large magnification droplet image group respectively; based on the conditional GAN ​​network, taking the small magnification droplet image group as input, generating a realistic magnified image, taking the large magnification droplet image group as the identification object, judging whether the generated realistic magnified image is consistent with the large magnification droplet image group, and finally establishing an association network model between the different magnification droplet images, so that the realistic magnified image is consistent with the large magnification droplet image group; All droplets in the image of the small magnification array are segmented, extracted and saved, and then imported into the established association network model of the different magnification droplet images to generate realistic magnified droplet images; Construct a ResNet neural network, import calibration images from the calibration database into the network, and establish an accurate recognition network for droplet particle size; Importing the realistic magnified droplet image into the precise recognition network to obtain particle size information of each droplet in the image of the small magnification image group; The cloud and fog field incoming flow parameters LWC and MVD are calculated based on the particle size information.

4. The high-precision dual-field-of-view wide-area particle size measurement method for high-speed cloud and fog fields according to claim 3 is characterized in that: In step (4), the particle size information is used to calculate the cloud and fog field flow parameter LWC as follows: （1） Where, V CSA The camera capture area, The first i The volume of the identified droplets; （2） Where, D max is the maximum droplet size in the field of view, S CSA is the field of view area perceived by the camera pixels and magnification; （3） Where, D i For the i The particle size of the detected droplets.

5. The high-precision dual-field-of-view wide-area particle size measurement method for high-speed cloud and fog fields according to claim 3 is characterized in that: In step (4), the particle size information is used to calculate the cloud field flow parameter MVD: （4） Where n is the total number of droplets in the field of view.

6. The high-precision dual-field-of-view wide-area particle size measurement method for high-speed cloud and fog fields according to claim 3 is characterized in that: In step (1), the test area at the small magnification end reaches 1mm 3 above.

7. The high-precision dual-field-of-view wide-area particle size measurement method for high-speed cloud and fog fields according to claim 3 is characterized in that: In step (1), the test particle size resolution at the large magnification end reaches below 1 μm.

Images