Optimize Image Quality#

This chapter provides guidelines to check and optimize the quality of images generated by the RadarImager. To use these guidelines, you should already be able to create images that show the object of interest.

The quality of captured images depends heavily on correct alignment and accurately determined object speed. Follow these guidelines to recognize and optimize image quality.

Strategy#

It is best to align the RadarImager properly using measuring tools and determine the object speed precisely. Sometimes this is not possible, or there are errors using tools to measure speed and alignment. It is possible to recognize and optimize alignment and speed errors based on this guideline. For fine-tuning, a trial and error approach can be used to improve results until the image quality is suitable.

Follow these steps to verify and optimize image quality:

• Mount the RadarImager so it is well aligned with the object movement direction. Image quality optimization requires correct alignment and speed determination to a certain extent.
• Determine the object speed as precisely as possible and configure the objectSpeedX parameter accordingly.
• Use a reference object to create an image and find the correct layer of interest.
• If you cannot see the reference object, refer to Typical Config Errors and try to find the object within the image.
• Compare the acquired image with a reference object image below and try to identify the error source that influences the image quality most:
  • Object speed deviation
  • RadarImager XY Orientation
  • RadarImager XZ Orientation
• Based on the identified error source, optimize the RadarImager orientation or the object speed using measurement approaches or trial and error. Repeat this until the image quality matches the reference object images. Optimizing only one error source at a time will help to identify improvements more easily.

Reference Objects#

To investigate and optimize image quality, use a reference object. A corner reflector is particularly suitable for this purpose. However, simple objects like a pair of metal scissors can also be very useful for optimization.

Corner Reflector#

A corner reflector is a device that reflects radar signals back to the source, regardless of the angle of incidence. It consists of three mutually perpendicular, intersecting flat surfaces, which cause the incoming radar waves to be reflected back in the direction they came from. Therefore, a corner reflector behaves like a point scatterer with high reflectivity for the RadarImager. An optimized image should show a nearly perfect point scatterer at the corner reflector as shown below.

To find the correct layer for the corner reflector using the layerSelection.indexes parameter, you may notice the triangle structure of the corner reflector in a higher layer. Try navigating through the layers to find the layer with the highest point scatterer reflections. Using the globalScale option can help to find the layer with the highest reflection.

globalScale = false

globalScale = true

These GIFs show the effect of the globalScale parameter in finding the correct layer of interest for using the corner reflector.

Scissors#

A pair of scissors or a similar object with a simple metallic surface and clear shape can also be helpful to optimize image quality.

Possible Error Sources#

There are three main error sources that can affect the image quality of the RadarImager. Have a look at the images below to identify the error sources.

Object Speed Deviation#

For image quality, it is crucial to know the correct object speed and set the objectSpeedX parameter on the RadarImager. The speed should be determined using external measuring tools, but if precise measurement is not possible, it can be optimized using a reference object and trial and error.

These GIFs show the effect of varying object speed in 2 mm/s steps.

A speed error can be identified by the smearing or distortion of the image in the direction of movement. The point scatterer will become an elongated oval shape or a line with increasing error. The scissors will also be distorted in the horizontal direction. Additionally, a significant speed error can cause ghost effects in the images, where the object appears multiple times. These ghost objects are usually not as bright as the actual object. Ghost objects can be distinguished from other error types as they do not lie on a perfectly vertical line but are slightly tilted.

It is possible to determine whether the speed is too high or low based on horizontal stretching or compressing of objects. However, depending on the reference object, this can be difficult, and a trial and error approach might be necessary.

RadarImager XY Orientation#

Refer to the overview picture for the RadarImager coordinate system.

If the XY orientation is not correct, the object will create smeared ghost images in the vertical direction. The effect of ghost images will increase with the angle of orientation mismatch to the point where the actual object is smudged.

These GIFs show a varying XY orientation error.

Objects affected by this orientation error tend to twist and lose their original shape.

It is not possible to determine the direction of orientation mismatch from the image, so a trial and error approach might be necessary.

RadarImager XZ Orientation#

Refer to the overview picture for the RadarImager coordinate system.

If the XZ orientation is not correct, the object will create clear ghost images in the vertical direction. The effect of ghost images will increase with the angle of orientation mismatch to the point where the ghost image becomes brighter than the actual object. This will lead to the effect that the actual object appears to be a ghost image. A ghost image that becomes brighter than the actual object will not be as sharp as the actual object. Always verify that the object in the image is at the expected position and check that the orientation is not tilted.

These GIFs show a varying XZ orientation error.

The ghost images created by this orientation error will always appear in a perfect vertical line and will be quite sharp compared to other possible errors.

It is not possible to determine the direction of orientation mismatch from the image, so a trial and error approach might be necessary.

✃