Basics of the RadarImager and its specific use#
The RadarImager is a 3D imaging system that uses radar radiation from several radar sources to create an image of a target moving on e.g. a conveyor belt. Since radar radiation is in the GHz range, it is partly transparent for many material dielectrics, like plastics, papers, cartonnage as glass in the optical range. This allows e.g. quality checks through packaging and other tasks which need look through behavior and local evaluation.
In the same way that glass influences optical views although it is transparent, the target/packaging situation influences radar radiation. Understanding to a certain point the physical context of radar interaction enables the real performance of the RadarImager, which makes it important to have basic understanding of radar physics and geometrical dependencies of the target/packaging.
This radar knowledge allows to solve evaluation tasks in general, and develop evaluation strategies which are even more powerful.
How a radar scan works#
A scan is a constant movement of a target under the RadarImager. Either the target is moved or the RadarImager. The image creation process can only be done after a full scan of the target. No line by line image creation is possible. This is necessary to realize the three-dimensionality of the image.
The scan of an object is schematically shown below. While the object is passing the RadarImager,
the object is scanned continuously line by line (the 2nd dimension is not shown here, to keep it simple, but the RadarImager is actually a line of antennas).
Each active radar element of the RadarImager has in each scan moment a cone view down of around 40° to the target area.
Therefore, after the scan, measurements are available from several sides and from all angles up to 40°.
This means, there is an all-round scan of the object from all viewing directions that can be reached by the antennas of the RadarImager.
Each of these measurements corresponds to a radial distance measurement to the (target-) object.
These measurement data is then algorithmically reconstructed/focused into a 3D grid,
reconstructing the reflections on material transitions as image stack.
The illustration above would be the ideal situation (clear geometry, only one object, one material type,
no covering, ...).
We will derive in the next chapters, what happens if the target gets more complex.
Resolution#
The RadarImager has a lateral resolution (XY-direction) and a range resolution (z-direction) with different definitions.
Resolution lateral (XY-direction)#
The lateral resolution is derived from the operation frequency and the dedicated wavelength. Ideally, the RadarImager can resolve within one image position/layer with approx. 1.7 mm at 60 GHz operation frequency, if the radar wave/beam hits the target object from all sides without disturbance. This means that it can separate two objects as brightness information if they are at least 1.7 mm apart from each other. If more geometric details are wished to be evaluated than just two points of light, a larger detail size of the object is required then the resolution limit.
Resolution limitation in depth (Z-direction)#
The range/depth resolution is derived from the operation bandwidth and legally limited. Our RadarImager has a depth resolution of 22 mm. This means that each layer selected from the image stack (e.g. double dash layer in the graphic below) is a collection of information from the surrounding approx. +-11 mm, with decreasing influence towards the edges. Therefore, if a target object and its packaging are closer together, than 22mm, then information overlaps in the layer display.
Three image cuts of a pan in a package. To a dominant level the layers have separate information, but partially the information still overlaps between the layers.
Influence of intermediate materials in front of the target object#
Fraction
Any dielectric material that lies within the scanning area in front of the actual target object, for example the packaging, manipulates the actual radar beam in terms of angle, time of flight and energy transmission, so the structure underneath is always a bit blurred by an additional material. The thinner this material is (carton, plastic housing or foils are typical), the less this manipulation is. This is why objects in air-filled bodies (packaging) can always be reconstructed more accurately than in solid materials, such as injection-molded parts with targets inserts.
Explanation: The rays of the RadarImager are reaching the (target-) object – dotted lines. The packaging is made of carton shown as outer box. The packaging is filled with air. The dashed grey lines are showing the exemplary local shift of the rays (black dashed line) caused by the packaging or nearby volume material, leading to deviation in target scans.
If the packaging is geometrically more extreme than just rectangular, then further effects are added. (see Influence of the geometry on the RadarImager)
Partial transparency
Each reflection (dielectric material transition/change along the ray propagation) partially reduces the energy and therefore the potential information of the radar wave for the deeper reflection layers, causing them to lose resolution quality. In extreme cases, they drop to noise level.
Special case: Metal objects are always full reflectors (clearly visible even with small structures), but they hide the area below the metal, because no beam/wave is getting there
Views of the radar images#
As explained at the beginning of this chapter, the RadarImager receives its previously transmitted radar waves/beams reflected from the target object. The complex radar image is then calculated by an algorithm, which contains both amplitude and phase information. The amplitude view and the phase view are two different methods of interpreting and displaying this data. In the settings (e.g. on the RadarImager software), you can choose between amplitude (absolute), phase display and a combination of both. This is explained in more detail in the following sub-chapters.
Magnitude (absolute) view#
When we talk about the magnitude view in radar, we mean the view of the strength (amplitude) of the returning waves from the target object that the radar imager has previously emitted. In a radar image, this information is displayed as brightness:
- Bright areas: Strong reflection (high amplitude)
- Dark areas: Weak (or no) reflection (low amplitude)
Here you can see the carton base/bottom of the mini pizzas. The carton base appears brighter than the mini-pizzas themselves, that means the carton base has a stronger reflection (higher amplitude) than the mini-pizzas, which reflect weakly (or barely at all).
Phase view#
The phase view (typically combined with the magnitude view as brightness information, as we are typically interested in the phase in the area of existing reflections, keeping all other regions black) is a fine resolution of the distance within a layer image and can resolve the distance more precisely within the wavelength range of approx. 2 mm. To put it simply: If, for example, two objects are almost the same size, # the phase display can be used to show very well which object is the larger one and therefore has a smaller distance to the radar. With the colormap (see below), the different heights are displayed in a different color and thus the difference is visible. Additionally, the phase of a wave is influenced by the strength and kind of transparency of a dielectric material.
However, the colors are repeated approximately every phase of 2pi (in distance around 2 mm) in range direction, as can be seen on the slanted pan handle. In practice, the phase view is used to detect very small changes in the position of an object like surface changes and for material signatures evaluation. It has to be considered in context with the real target, if a phase information delivers real additional value or not, since distance and material influence can overlap each other and can not separated during evaluation any more.
Roughly speaking, the color map of the phase display currently resolves in 6 colors.
In case of pure distance evaluation: two adjacent colors, e.g. red to purple or blue to green, mean approximately a height difference of 2 mm/6 = 0.33 mm. However, this should not be used for quantitative measurement, but only for qualitative evaluation. Uneven surfaces, as is typical with cardboard and thin plastics due to surface tension, also result in a change in color. Height comparison measurements over the phase should therefore ideally always be made directly on the object without packaging. The surface of the object must also be level/even.
Summary of the view variants#
- Magnitude view: Easier to understand, shows the "brightness" or reflectivity of the objects. (Often used to recognize physical properties of objects).
- Phase view: Provides detailed information about the transparency or exact position of objects. (Used in specialized applications such as structural integrity).
Influence of the geometry on the RadarImager#
The following instructions generally apply to geometries of objects. In other words, for packaging and the target objects.
Aerial constraints#
It is always that reflection information measured, which returns along the same path back as it was emitted (dotted arrows). In the case of a straight, flat surface, the reflection back is therefore less for a single scan to the wide angled sides, as the scattering from the object decreases, but due to the movement of the object there are always sufficient reflection fractions from other measurements. The worse resolution measurements add with other line measurement to even better resolutions. So all reflection measurements lead to improvement. In case of rough surfaces, the basic principle is the same, but angled reflection might lead to higher scattering and therefore higher signal strengths back to the antenna. This surfaces might be better recognizable in brightness.
Round geometry#
If the surface of the target is round, there are less and less reflection paths that lead from the sphere directly back to the antenna, the more the radar rays aren’t rectangular any more to the tangential level of the spherical structure. Therefore, round objects always shrink to small light spots in the case of spheres and to spot lines in the case of round cylinders after image reconstruction. However, the larger the radius of the object, the more of the object can be reconstructed in the image, because a large radius keeps longer almost rectangularity.
