Dynamic Image Analysis: Principles, Data Quality, and Applications

Dynamic image analysis relies on specialized instruments called dynamic image analyzers or particle analyzers. These instruments are designed to capture high-speed images of particles in motion. Key components of a typical dynamic image analyzer include:

• High-speed camera: A high-speed camera with a fast frame rate is used to record a sequence of images. These cameras can typically capture hundreds of images per second, enabling the observation of rapid particle movements.
• Lighting system: Adequate illumination is critical for capturing clear images of particles. The lighting system provides uniform and consistent illumination to ensure accurate analysis.
• Sample chamber: The sample is placed in a chamber where it is continuously agitated or dispersed. Various mechanisms, such as vibration, airflow, or fluid circulation, can be employed to keep the particles in motion.
• Optics: The optical system includes lenses and filters that focus and filter the light before it reaches the camera sensor.

As a technique, DIA shares certain similarities with both traditional light optical microscopy (LOM) and static image analysis (SIA). DIA also uses light and optics but differs from LOM and SIA as a) particles are in motion and constantly exchanged during the measurement, b) particles are randomly oriented towards the camera, and c) particles are detected and analyzed automatically. For these reasons, DIA can include very large numbers of particles in each single measurement, which is necessary to achieve good statistical quality of the resulting distributions (Figure 1). 

Once dispersed, the particles travel through the measurement zone using gravity, air pressure, or in a moving carrier liquid. Backlight illumination (shadow projection) is used to produce greyscale images of the particles. These projections are converted into binary images to determine the particles' contours, which can then be used for shape and size analysis.

DIA instruments have to be calibrated to convert pixels into SI length units (e.g., µm). For the calibration, a certified, static target with structures of known size is used. In the case of Litesizer DIA, the distance between the center of the dots is determined, and the mean distance is used to calculate the corresponding value, typically expressed in micrometers (Figure 3). 

For the validation of the measurement system, moving particles of a certified reference material with a known diameter have to be used.

According to ISO 13322-2, at least three pixels are required, aligned in one dimension, for acceptable sizing. At least nine pixels are required in one dimension for reliable shape analysis. The maximum particle size should be limited to one-third of the shortest side of the field of view. Exceeding this limit will significantly increase the risk of the particle’s image touching one of the measurement frame’s edges (Figure 4). It means that particles smaller than one-third the size of the shorter side of the field of view can be correctly measured.

If all particle images that appear in the frame are accepted for measurement, the accuracy of the measurement will be impaired because some particle images will inevitably be cut by the frame’s edges. To overcome this, particle images that are touching the edges must be excluded from the analysis (Figure 5). However, the probability that a particle’s image touches the edges increases with its size, also potentially skewing the distribution. To counter this effect, a statistical correction must be applied by the algorithm generating the particle size distribution.

The particles are imaged with maximum sharpness of their edges if they are positioned in the so-called object plane. However, the images can be accepted for analysis if the edges have a minimum sharpness. This threshold sharpness defines a depth of field where particles are imaged sharply enough (Figure 6). Furthermore, this depth of field defines a measurement volume that includes all measured particles. Particles outside of the depth of field are too blurred for analysis and have to be excluded from the measurement.

The exposure time of the imaging system is usually defined by the light pulse length (strobe light). The pulse length has to be short enough to avoid significant motion blur, even for the fastest particles (Figure 7). 

For Litesizer DIA, the opening time of the shutter is one millisecond, and the exposure time (illumination time) is around 100 ns (Figure 8). 

DIA enables the measurement of a wide range of parameters related to particles or objects within the sample. These parameters provide detailed insights into the sample's characteristics and behavior. Bulk properties, such as density, flowability, ballistics, and many others, are related not only to particle size but also to particle shape. 

Multiple size and shape parameters can be measured by DIA, and the most commonly used ones are presented in Table 1 and Table 2. 

A single universal parameter applicable to every application does not exist, so the parameters most relevant to a specific sample must be selected carefully. Systems offering multiple, different size and shape parameters and maximum flexibility and customization are therefore a must. 

Each particle can be categorized by multiple parameters at the same time, creating a multi-dimensional room. Nevertheless, the most practical approach for a user is taking two parameters, e.g., sphericity vs. particle size, to obtain a matrix. This allows us to quickly get an overview of the particles contained in the sample. The four corners show small spherical, small elongated, large spherical, and large elongated particles. Between the outermost edges and corners, the parameters vary gradually, showing also particles that do not represent extreme cases. 

In theory, a third filtering parameter could be added, e.g., irregularity, resulting in a 3D cube, cuboid, or other 3D structure. One edge could represent small elongated irregular particles, another one large elongated regular particles, and then small spherical irregular particles, etc. Adding a fourth parameter, e.g., convexity, would result in a 4D structure, which is already difficult to represent .

The sharpness of a particle's contour in an image is determined by the gradient of the greyscale value in the direction perpendicular to the contour, averaged over the entire contour. For a particle that is sharply imaged (b), the greyscale value changes rapidly from high (bright) to low (dark) as one moves from the background toward the particle's center. Consequently, the gradient of the greyscale value and the sharpness parameter are both high. In contrast, for a blurred image (a), the gradient of the greyscale value and the sharpness are lower. Sharpness is not an inherent property of the particle itself, but rather it depends on the imaging optics. Specifically, it is influenced by the relative position of the particle to the depth of field of the imaging optics. This parameter is useful for filtering out blurred particle images.

This parameter indicates the contrast of a particle image by measuring the difference in greyscale value (Y) between the background near the particle (Y_background) and the area within the particle (Y_particle). The contrast value ranges from 0 to 1, where 1 represents a completely black particle.

$\text{Contrast} = \frac{Y_{\text{background}} - Y_{\text{particle}}}{Y_{\text{background}}}$

This parameter is particularly useful when analyzing fine features on particle surfaces is important. Additionally, it is valuable for selecting photos for reporting or scientific publication.

The quality of cement is strongly dependent upon the size and shape of its particles, which affect its surface area, compression strength, and curing time. Particles that are too fine cause an exothermal setting in the final product, while particles that are too large do not fully hydrate. Cement flowability and water demand may vary drastically between regular (spherical) and irregular particles. In fact, more regular particles hydrate faster. 

An exemplary result for a cement sample is shown in Figure 9. The size distribution based on the x_A size model is monomodal and broad, with the major part of the population ranging from 5 μm to 60 μm (Figure 9A). The aspect ratio displayed in Figure 9B shows that most of the particles have an aspect ratio close to 1, indicating a spherical shape.

Abrasives include different types of materials, such as metals, ceramics, and polymers. They are used to polish and smooth the surface of metallic, plastic, or ceramic components. Their hardness, size, and shape all have an impact on the ultimate abrasive performance. Besides the aspect ratio, the shape parameter ‘circularity’ is of particular interest for abrasives. As displayed in Figure 10A, the size parameters as well as the aspect ratio/x_A overview are by default shown after the measurement. In the analysis window, the circularity distribution can also be listed and evaluated. Interestingly, the circularity distribution (Figure 10B) shows that, even though the four abrasive samples tested have very different particle sizes, they all have a very similar shape distribution. 

Most metal-based additive manufacturing (or 3D printing) technologies rely on metallic powders as a feedstock material. Since the quality of the powder used directly influences the quality of printed parts and printing speed, it is crucial to monitor its properties. The influence of the powder quality can be the most prominent in powder bed techniques, where the powder should have a high packing density and form smooth layers, usually in the range of 25 µm to 100 µm. Ultrasonic atomization is a liquid-to-solid process for the production of metallic powder, which uses ultrasonic vibrations to create the powders. The controlled atmosphere inside the atomization chamber will have an impact not only on the particle size but also on the shape. Metallic additive manufacturing expects a feedstock consisting of perfectly spherical particles. Elongated particles are difficult to separate after production and can therefore introduce irregularities into the powder bed. Figure 11 shows how the optimized conditions of the atomization process used for sample B increase the sphericity (aspect ratio close to 1) and reduce the characteristic flaky oxidation layer around the particle.