Spherical particle size measurement method based on interference imaging technology and Unit + +

Abstract

The invention discloses a spherical particle size measurement method based on an interference imaging technology and Unit + +. A spherical particle interference fringe pattern is obtained through an interference particle imaging system, and a simulation program is written to obtain a simulation fringe pattern; corresponding label graphs are made according to particles of different sizes, and a data set is divided into a training set, a verification set and a test set; the network is improved, and seven residual modules and one up-sampling module are added into the network; hyper-parameters are adjusted, training and testing are carried out through the data set, and finally the optimal weight is obtained. The network reconstructs spherical particle images of different sizes, and connected domains of the images are measured to obtain particle size information. The result shows that the average relative error of the particle size is less than 0.002%, and the average detection rate is 53.38 fps. Real-time and high-precision measurement of the spherical particle size is realized, and technical support is provided for acquisition of cloud particle field information.

Description

Technical Field

[0001] The invention belongs to the field of image processing, and in particular relates to a spherical particle size measurement method based on interference imaging technology and Unet++. Background Art

[0002] Clouds are composed of spherical water droplets and irregular ice crystal particles. The measurement of cloud particle size is conducive to the realization of high-precision weather forecasts and thus the avoidance of meteorological disasters. It has important scientific significance for revealing the formation and development process of clouds, artificial rainfall, etc. Interferometric Particle Imaging (IPI) technology is a measurement method based on the distribution of scattered light from particles. This method studies the relationship between the particle diameter and its scattered light distribution through the Mie scattering theory, receives the scattered light of the particles through an optical imaging system, and then obtains size information. In the research on particle size measurement using interferometric particle imaging technology, the measurement method of the existing patent CN108593528B is mainly aimed at irregular particles, and cannot be applied to the size measurement of spherical particles. In addition, the process is relatively complicated, which is not conducive to real-time acquisition of cloud particle field information.

[0003] At present, convolutional neural networks have been applied to the field of particle measurement. The measurement method of the existing patent CN116559033A realizes the shape reconstruction of irregular particles, but lacks the analysis of spherical particles. The present invention combines IPI technology with convolutional neural networks, and proposes a spherical particle size measurement method based on IPI technology and Unet++, which can realize real-time and high-precision measurement of spherical particle size. The difficulties of the present invention include the need to collect a large amount of experimental data to meet network training, adjust network hyperparameters and structures to obtain optimal weights, and detect interference fringe patterns in extreme cases to achieve high-precision size measurement. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the above-mentioned prior art and to provide a spherical particle size measurement method based on interference imaging technology and Unet++, so as to realize real-time and high-precision particle size measurement and provide technical guarantee for the measurement of cloud particle field.

[0005] The technical solution adopted by the present invention is: a spherical particle size measurement method based on interference imaging technology and Unet++, which is carried out according to the following steps:

[0006] Step 1: Build an interference particle imaging system to obtain the interference fringe pattern of spherical particles. The interference particle imaging system includes: a laser, a spatial filter composed of an objective lens and a pinhole, a collimating lens, a lens group composed of a plano-convex cylindrical lens and a plano-concave cylindrical lens, a sample pool, an imaging lens, and a CCD camera; a 532nm laser, a spatial filter composed of an objective lens and a pinhole, a collimating lens, a lens group composed of a plano-convex cylindrical lens and a plano-concave cylindrical lens, and a sample pool containing polystyrene spherical particles are arranged from left to right in the horizontal direction, and an imaging lens and a CCD camera are arranged below; the system is used to obtain the interference fringe pattern of spherical particles and construct a data set; the diluted polystyrene spherical particles are placed in the sample pool, and after being irradiated by a sheet laser beam, the interference fringe pattern of the spherical particles is obtained at the CCD end;

[0007] Step 2: Prepare data sets, label maps, and divide data sets; according to the imaging system and simulation program in step 1, obtain experimental and simulated interference fringe map data sets respectively, and prepare label maps corresponding to spherical particles of different sizes, where the label content contains particle size information; divide the spherical particle interference fringe map data set into training set, validation set, and test set in a ratio of 7:2:1;

[0008] Step 3: Improve the network structure. First, add 7 residual modules to the Unet++ network, specifically in the encoding and decoding part of the Unet++ network, located after the double-layer convolution; second, add 1 upsampling module to the Unet++ network, specifically in the decoding part of the Unet++ network, located in the final output layer; this reduces the number of parameters and calculations, and speeds up particle measurement;

[0009] Step 4: Train the network to obtain the optimal weights. Use the data set completed in step 2 and the improved network in step 3 for training. The hyperparameters are set as follows: The initial learning rate is set to 10 -6 , the learning rate momentum is set to 0.9, and the weight decay coefficient is set to 10 -8 ; Input the training set divided in step 2 into the network training for 200 epochs to obtain the optimal weight.

[0010] Step 5: Detect the interference fringe pattern to obtain size information; construct a mapping relationship between particle size and reconstructed image, and use the optimal weight obtained through training to obtain reconstructed images of particles of different sizes in the test set. Measure the distance of the connected domain in the reconstructed image and obtain the particle size through the mapping relationship. The relative error of the measurement result is less than 0.002%, and the average detection rate is 53.38fps, realizing real-time and high-precision measurement of spherical particle size.

[0011] The laser emits parallel polarized light with a wavelength of 532nm and a maximum output power of 3W.

[0012] A CCD camera with a resolution of 2448×2048 and a pixel size of 3.45μm is used as the receiving device.

[0013] The particles to be tested are polystyrene spherical particles of 30 μm, 45 μm, 60 μm, and 90 μm, which are diluted and placed in the sample pool.

[0014] The present invention has the following technical effects:

[0015] First, real-time processing. Input the particle interference fringe pattern, and output the reconstructed image and measurement value. Compared with traditional image processing methods, this method has higher real-time performance.

[0016] Second, high precision. Experimental noise affects the accuracy of measurement, and this technology can overcome the impact of noise on dimensional measurement to a certain extent.

[0017] Third, it has strong generalization ability. The interference fringe circle in extreme cases will not affect the accuracy of size measurement. The model has strong generalization ability and can be applied to actual scenarios.

[0018] The present invention realizes the size measurement of spherical particles by combining interference imaging technology with Unet++ neural network, providing technical support for the acquisition of cloud particle field information. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the particle size measurement method of the present invention;

[0020] Figure 2 This is a schematic diagram of the structure of the interference particle imaging system, including: 1. laser, 2. microscope objective, 3. pinhole, 4. collimating lens, 5. plano-convex cylindrical lens, 6. plano-concave cylindrical lens, 7. sample cell, 8. imaging lens, 9. CCD camera.

[0021] Figure 3 Schematic diagram of interference fringe patterns and processing results, where: (a)-(f) are experimental interference fringe patterns of spherical particles of different sizes; (g)-(l) are size label diagrams of spherical particles; (m)-(r) are the horizontal and vertical distance result diagrams of the connected areas of the label diagram.

[0022] Figure 4 Graphs showing the size measurement results of spherical particles, where (a)-(d) are the size measurement results of spherical particles with diameters of 30 μm, 45 μm, 60 μm and 90 μm respectively. DETAILED DESCRIPTION

[0023] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0024] The process of the spherical particle size measurement method based on interference imaging technology and Unet++ is as follows: Figure 1 As shown, specifically:

[0025] Step 1: Build Figure 2 The interference particle imaging system shown in the figure is used to obtain the interference fringe diagram of spherical particles. The interference particle imaging system includes: a laser, a spatial filter composed of an objective lens and a pinhole, a collimating lens, a lens group composed of a plano-convex cylindrical lens and a plano-concave cylindrical lens, an imaging lens and a CCD camera; a 532nm laser 1, a spatial filter composed of an objective lens 2 and a pinhole 3, a collimating lens 4, a lens group composed of a plano-convex cylindrical lens 5 and a plano-concave cylindrical lens 6, and a sample pool 7 containing polystyrene spherical particles are arranged in sequence from left to right in the horizontal direction, and an imaging lens 8 and a CCD camera 9 are arranged in sequence below the sample pool 7; the system is used to obtain a data set of interference fringe diagrams of spherical particles, and the system parameters are set as follows: the object distance is set to 140.90mm, the image distance is 77.50mm, the scattering angle is 60.73°, and the magnification is 0.55. The particles to be tested are 30μm, 45μm, 60μm and 90μm polystyrene spherical particles, which are placed in a sample pool filled with deionized water; after being irradiated by a sheet laser beam, the interference fringe pattern of the spherical particles is finally obtained at the CCD end;

[0026] Step 2: Create data sets, label maps, and divide data sets. According to the imaging system and simulation program in step 1, obtain experimental and simulated interference fringe map data sets respectively. Create label maps corresponding to spherical particles of different sizes, and the label content contains particle size information; a total of 9984 data sets are obtained, including 3726 simulations and 6258 experiments, with a resolution of 256×256. Divide the spherical particle interference fringe map data set into training set, validation set, and test set in a ratio of 7:2:1;

[0027] Step 3: Improve the network structure. First, add 7 residual modules to the Unet++ network, specifically in the encoding and decoding parts of the Unet++ network, located after the double-layer convolution; second, add 1 upsampling module to the Unet++ network, specifically in the decoding part of the Unet++ network, located in the final output layer;

[0028] Step 4: Train the network to obtain the optimal weights. Use the data set completed in step 2 and the improved network in step 3 for training. The specific training steps are: first set the hyperparameters and set the initial learning rate to 10 -6 , the learning rate momentum is set to 0.9, and the weight decay coefficient is set to 10 -8 . Secondly, the training set divided in step 2 is input into the network for training for 200 epochs to obtain the optimal weight.

[0029] Step 5: Detect the interference fringe pattern and obtain the size information. Construct the mapping relationship between the particle size and the generated image, and use the optimal weights obtained through training to obtain the reconstructed images of particles of different sizes in the test set. Measure the distance of the connected domain in the reconstructed image, and obtain the particle size through the mapping relationship. Figure 3 As shown in (a) and (m), the reconstructed image connected domain obtained from the interference fringe pattern of a 30μm particle is 60 pixels. According to the mapping relationship, the number of pixels corresponds to twice the size value, so the size of the spherical particle is 30μm. The relative error of the measurement result is less than 0.002%, and the average detection rate is 53.38fps, achieving real-time and high-precision measurement of the size of spherical particles.

[0030] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0031] Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. Those skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

[0032] The above specific implementations of the present invention do not constitute a limitation on the protection scope of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for measuring the size of spherical particles based on interferometric imaging technology and Unet++, characterized in that: The method comprises the following steps: Step (1), constructing an interference particle imaging system to obtain interference fringe patterns of spherical particles of different sizes; the interference particle imaging system comprises: a laser, a spatial filter composed of an objective lens and a pinhole, a collimating lens, a lens group composed of a plano-convex cylindrical lens and a plano-concave cylindrical lens, a sample pool, an imaging lens and a CCD camera; a 532 nm laser, a spatial filter composed of an objective lens and a pinhole, a collimating lens, a lens group composed of a plano-convex cylindrical lens and a plano-concave cylindrical lens, and a sample pool are arranged in sequence from left to right in the horizontal direction, and an imaging lens and a CCD camera are arranged in sequence behind the sample pool; the interference fringe pattern of spherical particles is obtained by using the system and a data set is constructed: diluted polystyrene spherical particles are placed in the sample pool as particles to be measured, and after being irradiated by a sheet laser beam, an interference fringe pattern of the spherical particles is obtained at the CCD end; Step (2), preparing a data set, a label map and dividing the data set; obtaining experimental and simulated interference fringe map data sets according to the imaging system of step (1) and the prepared simulation program, and preparing label maps corresponding to spherical particles of different sizes, wherein the label content includes particle size information; dividing the spherical particle interference fringe map data set into a training set, a validation set and a test set in a ratio of 7:2:1; Step (3), improve the network structure; first, add 7 residual modules to the Unet++ network, specifically in the following locations: the Unet++ network encoding and decoding part, located after the double-layer convolution; second, add 1 upsampling module to the Unet++ network, specifically in the following location: the Unet++ network decoding part, located in the final output layer; Step (4), train the network to obtain the optimal weights; use the data set completed in step (2) and the improved network in step (3) for training. The specific training steps are as follows: first set the hyperparameters: the initial learning rate is set to 10 -6 , the learning rate momentum is set to 0.9, and the weight decay coefficient is set to 10 -8 Next, the training set divided in step (2) is input into the network for 200 rounds of training to obtain the optimal weights. Step (5), detect the interference fringe pattern to obtain size information; construct a mapping relationship between particle size and reconstructed image, and use the optimal weight obtained through training to obtain reconstructed images of particles of different sizes in the test set. Measure the distance of the connected domain in the reconstructed image, and obtain the particle size through the mapping relationship. The relative error of the measurement result is less than 0.002%, and the average detection rate is 53.38fps, realizing real-time and high-precision measurement of spherical particle size.

Images