|
|
Detect and follow particles
|
|
|
=================================
|
|
|
Most experiments study a model 2 dimensional (2D) granular system composed of circular discs as grains. While in cases of studying particle shape effect, discs are replaced by other shapes like ellipses, polygons and crosses. With a digital camera, the first and one of the most important information that can be obtained is particle positions. Particle detection and tracking alone provide rich information such as particle configuration and flow field, and are necessary for further measurements like force-bearing contact detection and contact force calculation. In this session, a detailed description for detecting particle positions and orientations and tracking particles in high resolution images will be summarized below. In addition, in images with resolution that is not high enough to precisely detect particle positions, another method (PIV) will be introduced.
|
|
|
Most experiments study a model 2 dimensional (2D) granular system composed of circular discs as grains. With a digital camera, the first and one of the most important information that can be obtained is particle positions. Particle detection and tracking alone provide rich information such as particle configuration and flow field, and are necessary for further measurements like force-bearing contact detection and contact force calculation. In this session, a detailed description for detecting particle positions and orientations and tracking particles in high resolution images will be summarized below. In addition, in images with resolution that is not high enough to precisely detect particle positions, another method (PIV) will be introduced.
|
|
|
|
|
|
|
|
|
## 1 Position detection and tracking
|
|
|
## 1 Position and orientation detection
|
|
|
|
|
|
### 1.1 Detection of the particles (discs) from white light imaging
|
|
|
### 1.1 Detection of the particles centers from white light imaging
|
|
|
|
|
|
For particles of disc shape, a picture with white light transmitted through particles is usually taken to record their positions. The general idea to detect particles then is to produce a high contrast or a binary image distinguishing between areas occupied by particles and those by background. The procedure can be summarized as follows:
|
|
|
|
... | ... | @@ -20,11 +20,11 @@ The procedure described above with results at each step is illustrated in the fi |
|
|
|
|
|
The convolution method is fast in computation, but requires high contrast image and has low tolerance for light noises inside of particles. By contrast, the circular Hough transform method has high tolerance for light noises inside of particles and are robust for particles that are deformed from a circular shape, however, is computationally expensive.
|
|
|
|
|
|
### 1.2 Detection of the particles from UV light imaging
|
|
|
|
|
|
Though used most commonly in granular studies, discs are different from shapes of grains in reality. Hence there have been numerous studies on granular systems with different shapes other than discs, e.g., ellipses, polygons and star-shape-like particles. Here we choose star particles as an example to illustrate how to detect them from UV light imaging, instead of white light imaging.
|
|
|
### 1.2 Detection of the particles orientation UV light imaging
|
|
|
|
|
|
Due to particles being frictional in most experiments, particles will usually rotate if the granular system undergoes deformation. Hence tracking the rotation of particles is also important, which requires the knowledge of particle orientations. To achieve this, one can draw a bar with UV ink on the surface of the particle. This bar will leave no mark under white light, thus having no interference with particle center detection process as introduced above. Once illuminated by UV light with white light off, only the UV bars are visible and can be recorded by camera. Due to high contrast in this situation (blue for UV bars and black for background), the image can be easily binarized with [Matlab function imbinarize](https://www.mathworks.com/help/images/ref/imbinarize.html) so that the UV bars are 1 and the rest of the particle is 0. With previously determined particle center and diameter, each bar can be associated with its corresponding particle. With the positions of all the pixels inside the bar, a least squares fit with the minimized perpendicular offsets reveals a linear function between the x and y positions in each UV bar. The slope of the line gives the orientation associated with each particle. An example illustrating this procedure is shown below.
|
|
|
|
|
|
![Fig_DisksOri](uploads/9bfdd1555c24d731e6e00793149115df/Fig_DisksOri.png)
|
|
|
|
|
|
## 2 Tracking
|
|
|
|
... | ... | |