A method for characterizing particle motion intensity based on particle mask-constrained frame difference

Abstract

This invention discloses a method for characterizing particle motion intensity based on particle mask-constrained frame difference, belonging to the technical field of particle material image processing and experimental-simulation comparative analysis. The method selects two frames from a particle motion image sequence, one at the target time and one at a subsequent time, performs identical ROI cropping, grayscale conversion, and particle region extraction, and takes the union of the two particle masks as the effective calculation region. Within this region, the grayscale difference amplitude is calculated, and after a first percentile robust normalization, non-particle region invalidation, nearest neighbor filling and smoothing, and a second percentile robust normalization, the particle motion intensity field is obtained. This method does not require single-particle tracking or traditional cross-correlation velocity vector solving, and can stably characterize the motion region and intensity distribution in dense, wet, or occluded particle systems. It can also be compared temporally with discrete element simulation velocity fields, contact number fields, and force chain fields.

Description

Technical Field

[0001] This invention belongs to the technical field of particulate material image processing, particle motion state characterization and experimental-simulation comparative analysis, specifically involving a particle motion intensity characterization method based on particle mask constrained frame difference. Background Technology

[0002] Particulate materials undergo complex discontinuous motions during unloading, accumulation, collapse, shearing, and mixing. Their macroscopic flow morphology is influenced by particle contact, friction, cohesion, liquid content, and boundary conditions. To evaluate the rationality of particulate material parameters and discrete element simulation models, it is necessary to stably obtain the particle motion region and the distribution of relative motion intensity from experimental images and compare them with the simulated velocity field, contact number field, or force chain field under the same time sequence.

[0003] Traditional particle image velocimetry techniques typically calculate displacement vectors based on cross-correlation of tracer particle images or particle tracking methods, and then convert the velocity field by time intervals. However, in dense particle, wet particle, or strongly occluded particle systems, overlapping particle boundaries, local reflections, brightness fluctuations, and insufficient particle texture can lead to unstable particle recognition and matching. Near the discharge port, rapid particle rearrangement further amplifies the uncertainty of traditional vector solutions.

[0004] Therefore, a method for characterizing particle motion intensity is needed that does not require single-particle tracking and traditional cross-correlation vector solving as necessary conditions. This method should be able to define the effective particle region in complex experimental images, suppress background and reflective interference, and output a motion intensity field and diagnostic parameters that correspond to the simulation results. Summary of the Invention

[0005] The purpose of this invention is to provide a particle motion intensity characterization method based on particle mask constrained frame difference, in order to solve the problems of existing particle image velocimetry or motion characterization methods in dense particle, wet particle and strongly occluded particle scenes, such as difficulty in particle identification, unstable vector matching, significant noise interference, and difficulty in stably comparing experimental images with simulation results.

[0006] To achieve the above objectives, the present invention employs the following technical solution: acquiring continuous experimental images of particulate material motion; selecting two adjacent or near-adjacent frames within the same region of interest and performing grayscale processing; extracting the particle regions from the two frames based on the grayscale difference between the particles and the background, and constructing an effective particle mask through mask union; calculating the difference amplitude of the two grayscale images within the effective particle mask, and performing robust normalization using a first set of percentile thresholds; setting non-particle regions as invalid regions, and filling and smoothing the motion intensity values ​​within the effective regions; further performing secondary normalization on the smoothed continuous motion intensity map using a second set of percentile thresholds, ultimately obtaining the particle motion intensity field.

[0007] Compared with existing technologies, this invention has the following advantages: First, by effectively defining the computational area using a particle mask, it reduces the interference of background, borders, scales, transparent container walls, and particle-free areas on the frame difference results; Second, by using robust two-level percentile normalization, it reduces the impact of abnormal brightness changes, local reflections, or noise on the overall motion intensity scale; Third, it does not require single-particle tracking or cross-correlation velocity vector solving as a necessary condition, making it more suitable for the unloading process of dense particles and wet particles; Fourth, the output particle motion intensity field can be compared temporally with the discrete element simulation velocity field, contact number field, or force chain field, providing a unified data basis for particle parameter calibration and model verification. Attached Figure Description

[0008] Figure 1 This is a flowchart of the particle motion intensity characterization method of the present invention.

[0009] Figure 2 This diagram illustrates the particle mask generation and union constraint process of the present invention.

[0010] Figure 3 This diagram illustrates the frame differencing, padding smoothing, and two-level percentile robust normalization processes of this invention.

[0011] Figure 4 This is a comparison chart of the particle motion intensity field output and the experimental-simulation results of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] like Figure 1 As shown, this embodiment provides a method for characterizing particle motion intensity based on particle mask-constrained frame difference. First, an experimental image sequence of the particle material's motion process is read. The first frame image is selected based on the target time t, and the second frame image corresponding to the subsequent time t+Δt is determined based on the video frame rate fps. Here, Δt can be taken as min(1 / fps, 0.04s) to balance inter-frame particle displacement and image difference stability.

[0014] Subsequently, the first and second frames were cropped using the same region of interest to remove borders, rulers, irrelevant backgrounds, and grainless areas. The two cropped frames were then converted to grayscale using the formula G = 0.299R + 0.587Gc + 0.114B, where Gc represents the green channel grayscale value to avoid confusion with the grayscale image G.

[0015] like Figure 2 As shown, grain mask generation is performed on the first grayscale image Gt and the second grayscale image Gt+Δt, respectively. Specifically, a grayscale percentile threshold is calculated within the effective statistical region of the image, and pixels with grayscale values ​​below the threshold are identified as grain pixels; at the same time, image edge regions and boundary regions that are obviously not grain regions are excluded. A first grain mask Mt is obtained from the first grayscale image, and a second grain mask Mt+Δt is obtained from the second grayscale image. The union of the two is taken to obtain the effective grain mask M, i.e., M=Mt∪Mt+Δt. The effective grain mask M defines the effective region for subsequent difference calculation, percentile statistics, and motion intensity output.

[0016] like Figure 3 As shown, under the constraint of the effective particle mask M, the gray-level difference amplitude map D between two gray-level images is calculated, i.e., D = |Gt + Δt − Gt|. Within the particle region defined by the effective particle mask M, the first lower percentile and the first higher percentile of the difference amplitude map D are calculated. In a preferred embodiment, the first lower percentile is the 20th percentile, and the first higher percentile is the 99th percentile, denoted as b and h respectively. The first motion intensity map A0 is obtained according to A0 = clip((D − b) / (h − b), 0, 1) · vmax.

[0017] Next, the non-particle regions outside the effective particle mask M are set to null or invalid values. The first motion intensity map A0 is then filled with nearest-neighbor data and smoothed using Gaussian smoothing, box smoothing, or weighted field-based smoothing to obtain the continuous motion intensity map A1. Further, the second lower percentile and second higher percentile values ​​of the continuous motion intensity map A1 are calculated within the particle region defined by the effective particle mask M. In a preferred embodiment, the second lower percentile is the 48th percentile, and the second higher percentile is the 99.4th percentile, denoted as f and c respectively. The particle motion intensity field A is obtained according to A = clip((A1−f) / (c−f),0,1)·vmax. After normalization, the non-particle regions are reset to invalid regions to ensure that the output only represents the particle region.

[0018] like Figure 4As shown, the particle motion intensity field A can be used to display the relative motion intensity distribution of particulate materials during unloading, collapse, accumulation, shearing, or mixing processes. For experimental images under different liquid contents, different particle materials, or different contact model conditions, the same region of interest, time interval, percentile parameter, and display upper limit can be used to obtain a comparable particle motion intensity field sequence. This sequence can be output correspondingly to the velocity field, contact number field, or force chain field obtained from discrete element simulation at the same time nodes, thus forming a spatial pattern comparison between experimental images and simulation results.

Claims

1. A method for characterizing particle motion intensity based on particle mask-constrained frame difference, characterized in that, The process includes the following steps: S1, acquiring an image sequence during the movement of particulate material, and determining the first frame image and the second frame image corresponding to the target time t and the subsequent time t+Δt; S2, performing the same region of interest cropping on the first frame image and the second frame image, and converting them into a first grayscale image Gt and a second grayscale image Gt+Δt respectively; S3, extracting the particulate region from the first grayscale image and the second grayscale image respectively, obtaining a first particulate mask Mt and a second particulate mask Mt+Δt, and taking the union of the two to obtain the effective particulate mask M; S4, calculate the grayscale difference amplitude map D=|Gt+Δt−Gt| within the area defined by the effective particle mask M; S5, perform a first robust normalization based on the low and high percentile values ​​of the grayscale difference amplitude map D within the effective particle mask M to obtain the first motion intensity map A0; S6, set the area outside the effective particle mask M as an invalid area, and fill and smooth the first motion intensity map A0 to obtain the continuous motion intensity map A1. S7. Based on the low and high percentile values ​​of the continuous motion intensity map A1 within the effective particle mask M, a second robust normalization is performed to obtain the particle motion intensity field A. S8, output the particle motion intensity field A, the corresponding effective particle mask M, time interval Δt, frame number and normalization parameter.

2. The method according to claim 1, characterized in that, The time interval Δt is determined by the image sequence frame rate fps, preferably Δt = min(1 / fps, 0.04s).

3. The method according to claim 1, characterized in that, The grayscale value is converted to G=0.299R+0.587Gc+0.114B, where R, Gc, and B are the grayscale values ​​of the red, green, and blue channels of the color image, respectively.

4. The method according to claim 1, characterized in that, The particle region extraction includes: calculating a gray-level percentile threshold within the effective statistical area of ​​the gray-level image, and determining pixels with gray-level values ​​lower than the threshold as particle pixels; the gray-level percentile threshold is jointly limited by the gray-level percentile value and the gray-level upper limit value, wherein the gray-level percentile value is preferably 42% to 46%, and the gray-level upper limit value is preferably 170 to 176.

5. The method according to claim 1, characterized in that, The effective particle mask M is M=Mt∪Mt+Δt, and the effective particle mask M is used to define the region for difference calculation, percentile statistics and motion intensity output.

6. The method according to claim 1, characterized in that, The first robust normalization includes: calculating the 20th percentile value b and the 99th percentile value h of the gray-scale difference amplitude map D within the effective particle mask M, and obtaining the first motion intensity map A0 according to A0=clip((D−b) / (h−b),0,1)·vmax, where vmax is the upper limit of display or output.

7. The method according to claim 1, characterized in that, The filling and smoothing include: setting non-particle regions outside the effective particle mask M to null or invalid marker values, performing nearest-neighbor filling on the motion intensity values ​​within the effective particle region, and obtaining a continuous motion intensity map A1 using Gaussian smoothing, box smoothing, or weighted field-based smoothing.

8. The method according to claim 1, characterized in that, The second robust normalization includes: calculating the 48th percentile value f and the 99.4th percentile value c of the continuous motion intensity map A1 within the effective particle mask M, obtaining the particle motion intensity field A according to A=clip((A1−f) / (c−f),0,1)·vmax, and resetting the non-particle region as an invalid region.

9. The method according to claim 1, characterized in that, The particle motion intensity field A is used to characterize the relative motion intensity distribution of particulate materials during unloading, accumulation, collapse, shearing or mixing processes, and is compared with the velocity field, contact number field or force chain field obtained by discrete element simulation at the same time node in spatial pattern.

10. A particle motion intensity characterization system or computer-readable storage medium, characterized in that, The system includes an image acquisition module, a frame selection module, an image preprocessing module, a grain mask generation module, a frame difference calculation module, a first normalization module, a fill smoothing module, a second normalization module, and a result output module; or the program stored on the computer-readable storage medium implements the method described in any one of claims 1 to 9 when executed by a processor.

Images