CN-122024002-A - Method for constructing measurement model of particle rotation speed in liquid swirling flow field and measurement method

CN122024002ACN 122024002 ACN122024002 ACN 122024002ACN-122024002-A

Abstract

The invention relates to a method for constructing a measuring model of the rotating speed of particles in a liquid swirling flow field and a measuring method, wherein a high-speed camera system and transparent or semitransparent test particles containing two inner cores are utilized to acquire an image sequence, a deep learning method is adopted to train a particle detection and classification model applicable to micron-sized particles, and a tracking algorithm is used to combine the particle detection and classification model to generate a particle tracking model; inputting the video to be measured into a constructed particle detection and classification model and a particle tracking model, acquiring basic data of the calculated rotating speed of the tested particles in the video to be measured based on a digital image processing technology, substituting the basic data into a rotating speed calculation formula, calculating the revolution speed and the rotation speed of the tested particles, and realizing the automatic measurement of the micron-sized particle rotating speed in the liquid rotating field.

Inventors

HE FENGQIN
LIU WEIQING
LIU QIBIN

Assignees

上海师范大学

Dates

Publication Date: 20260512
Application Date: 20251225

Claims (9)

1. The method for constructing the measurement model of the particle rotation speed in the liquid swirling flow field is characterized by comprising a particle detection and classification model and a particle tracking model, and comprises the following steps: step 1, acquiring an image sequence of two-dimensional movement of a test particle in a liquid swirling flow field through a high-speed camera system; step 2, carrying out data annotation on the image sequence by adopting data annotation software, and correspondingly generating an image sequence data set; Training the image sequence data set by adopting a deep learning algorithm to obtain a particle detection and classification model; And step 4, adopting a tracking algorithm and combining the particle detection and classification model to obtain a particle tracking model.
2. The method for constructing a measuring model of the rotational speed of particles in a liquid swirling flow field according to claim 1, wherein the outer shell of the test particles is transparent or semitransparent spherical, and two cores with symmetrical centers and equal diameters are arranged in the outer shell.
3. The method for constructing a measurement model of the rotational speed of particles in a liquid swirling flow field according to claim 1, wherein the high-speed camera system comprises two high-speed cameras which are orthogonally distributed.
4. The method for constructing a measurement model of the rotational speed of particles in a liquid swirling flow field according to claim 2, wherein the step 2 specifically includes performing data labeling on the image sequence by using data labeling software, labeling contents include separation and overlapping of projections of two cores of the test particles in the image sequence, and labeling information includes states of the projections and coordinate information of the test particles, so as to generate an image sequence data set.
5. The method according to claim 4, wherein the step 3 specifically includes using a deep learning model as a predetermined particle detection and classification model, using the image sequence data set as a training data set, and training the predetermined particle detection and classification model to generate the particle detection and classification model.
6. The method for constructing a measurement model of the rotational speed of particles in a liquid swirling flow field according to claim 5, wherein the step 4 specifically comprises inputting continuous frames in an image sequence into the particle detection and classification model, processing the continuous frames by the particle detection and classification model to generate a detection frame of each frame target, obtaining a detection result and a track state of a preamble frame, estimating the state of the corresponding target of the current frame according to the track of the preamble frame by using a Kalman filter to generate a prediction frame, calculating an overlapping area between the prediction frame and the detection frame of the current frame, constructing an association cost matrix based on the overlapping degree, performing global optimal matching on the prediction frame and the detection frame of the current frame by adopting a Hungary matching algorithm, and performing state updating and management according to the matching result to obtain a particle tracking model: Judging whether the overlapping area of the prediction frame and the detection frame of the current frame is larger than a preset threshold value, if so, judging that the matching is effective, using the detection frame of the current frame as an observation value, and updating the ID state information of the track of the preamble frame through the state correction process of a Kalman filter; If the track of the previous frame is not matched with any detection frame in the current frame, accumulating continuous unmatched frame numbers, judging that the target disappears when the frame numbers exceed a set value, and stopping tracking the ID of the track of the previous frame; If there is a detection frame which is not successfully matched with the track of any existing preceding frame in the current frame, the detection frame is judged to be a new target, a new track is initialized, a new track ID is allocated, and the tracking cycle of the next frame is included.
7. A model for measuring the rotational speed of particles in a liquid swirling flow field, which is characterized in that the model is constructed by adopting the method for constructing the model for measuring the rotational speed of particles in the liquid swirling flow field according to any one of claims 1 to 6.
8. A method for measuring the rotating speed of particles in a liquid rotating field is characterized in that a video to be measured is input into a particle detection and classification model and a particle tracking model constructed by the method for constructing a measuring model of the rotating speed of the particles in the liquid rotating field according to any one of claims 1 to 6, basic data of the rotating speed of the tested particles in the video to be measured is obtained based on a digital image processing technology, the basic data is substituted into a rotating speed calculation formula, and the revolution speed and the rotation speed of the tested particles are calculated, wherein the basic data comprises a migration distance, inner diameters of different heights of a cyclone, a margin and rotation frequency.
9. The method for measuring the rotational speed of particles in a liquid swirling flow field according to claim 8, wherein the video to be measured is input into the particle detection and classification model and the particle tracking model, the test particles appearing in the video to be measured are numbered, the state of the projection of the inner core in the test particles in different video frames and the coordinate information of the test particles in each numbered test particle are obtained, and the basic data are calculated.

Description

Method for constructing measurement model of particle rotation speed in liquid swirling flow field and measurement method Technical Field The invention relates to the field of liquid-solid two-phase flow testing, in particular to a method for constructing a measuring model of particle rotation speed in a liquid swirling flow field and a measuring method. Background The suspended particles do macroscopic displacement with the fluid in the fluid movement, and are stressed unevenly due to the velocity gradient of the flow field, so that the suspended particles generate rotary movement around the self axis. The rotation not only affects the motion characteristics of the particles, but also affects the surrounding solid-phase particle fields and flow fields, thereby affecting the efficiency and performance of the separation, desorption, absorption, trapping and other processes. The past studies have focused mainly on theoretical and experimental aspects, whereas most experimental studies have focused mainly on the autorotation movement of particles in high viscosity, simple shear flow, whereas relatively few studies have been conducted on the behavior of micron-sized particles in the swirling flow field. Measuring the rotation speed of particles is one of the key problems of research, and CN104049100A patent of a method and a device for testing the rotation speed of micron-sized particles in a liquid rotation field develops a method for testing the rotation speed of micron-sized particles in the liquid rotation field based on the combination of easily-identified particles and a high-speed camera system. However, the manual calculation of the rotational speed of the micron-sized particles by the method requires a great deal of labor cost and time cost, and is still a difficult problem to be solved for the automatic measurement of the rotational speed of the micron-sized particles in the liquid swirling field. Disclosure of Invention The invention aims to provide a method for constructing a measuring model of the rotating speed of particles in a liquid swirling flow field and a measuring method thereof, which solve the problems in the prior art. The technical scheme for solving the technical problems is as follows: The method for constructing the measurement model of the particle rotation speed in the liquid swirling flow field comprises a particle detection and classification model and a particle tracking model, and comprises the following steps: step 1, acquiring an image sequence of two-dimensional movement of a test particle in a liquid swirling flow field through a high-speed camera system; step 2, carrying out data annotation on the image sequence by adopting data annotation software, and correspondingly generating an image sequence data set; Training the image sequence data set by adopting a deep learning algorithm to obtain a particle detection and classification model; And step 4, adopting a tracking algorithm and combining the particle detection and classification model to obtain a particle tracking model. On the basis of the technical scheme, the invention can be improved as follows. Further, the outer shell of the test particle is transparent or semitransparent spherical, and two cores with symmetrical centers and equal diameters are arranged in the outer shell. Further, the high-speed camera system comprises two high-speed cameras which are distributed in an orthogonal mode. Further, the step 2 specifically includes performing data labeling on the image sequence by using data labeling software, wherein labeling content includes separation and overlapping of projections of two cores of the test particle in the image sequence, and labeling information includes a state of the projections and coordinate information of the test particle, so as to generate an image sequence data set. Further, the step 3 specifically includes taking a deep learning model as a preset particle detection and classification model, taking the image sequence dataset as a training dataset, and training the preset particle detection and classification model to generate the particle detection and classification model. Further, the step 4 specifically includes inputting continuous frames in an image sequence into the particle detection and classification model, processing the particle detection and classification model to generate a detection frame of each frame target, obtaining a detection result and a track state of a preamble frame, predicting the state of a target corresponding to a current frame according to the track of the preamble frame by using a Kalman filter to generate a prediction frame, calculating an overlapping region between the prediction frame and the detection frame of the current frame, constructing an association cost matrix based on the overlapping degree, performing global optimal matching on the prediction frame and the detection frame of the current frame by adopting a Hungary matching algorithm, and perform