RoadTest: Pi NoIR and Catch Santa Challenge
Author: gpolder
Creation date:
Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: NA
What were the biggest problems encountered?: NA
Detailed Review:
First of all I like to thank Element14 for giving me the opportunity to road test this nice bundle. The bundle was delivered in pieces, and some of the parts, like the camera and case were delayed. Nevertheless I had a lot of fun playing with the different components. For the road test we need to review only the camera. So what follows here is a description of my experiments with the camera.
The Pi NoIR (No InfraRed) is the second camera board made specifically for the Raspberry Pi. The only difference compared to the original camera is that the infrared filter has been removed.
Frederick Vandenbosch did already a nice review and showed some images of the camera and packaging. this review can be found here: http://www.element14.com/community/roadTestReviews/1638
I will not repeat his work, therefore in this review I will focus on the spectral aspects of the camera and describe some preliminary resolution tests.
The specifications of the camera can be found here: http://elinux.org/Rpi_Camera_Module
To look at these aspects I made some images of a Macbeth color reference chart and a USAF 1951 resolution test chart. Different types of illuminations and filters were used to get an impression of the spectral aspects. In first instance these charts were imaged fullscreen, but the distance to the camera was to short and focus blur was visible in the image. From the specs we see that the focus range is 1m to infinity. Therefore I placed the test objects 2m from the camera. The picture below gives an impression of the measurement setup.
The image below shows the color and resolution reference charts at 50% of their original resolution.
These images were cropped from the full image above. four types of illumination were used:
Two types of filters were used:
In the overview image below you can see the effect of the several combinations.
For instance the image fl_hal_ir_irf is the combination of fluorescence, halogen and the infrared floodlight, with the infrared cutoff filter before the camera. As expected, since the infrared from the floodlight and the halogen lamp is filtered by the cutoff filter this image is similar to fl, fl_hal_irf and hal_irf.
As you can see the images with infrared are rather pale. This is an indication that the red, green and blue pixels all are sensitive in the IR range. This can also be seen in the images with only infra red.
The images were taken with raspistill:
pi@rpicamera ~/test $ raspistill -n -t 1 -q 90 -awb off -ex auto -o dl.jpg
Since we don’t like automatic adaptations of the colors due to the different illumination spectra the auto white balance was set to off.
Exposure was set to auto and to prevent for compression artifacts, the jpeg quality was set at 90%.
A few words should be devoted to the blue filter. This filter is a so called ‘NGB’ or ‘infra blue’ filter.
It filters out the red part of the spectrum and can be used for plant health analysis. You can find a lot of information about this technique here: http://publiclab.org/wiki/near-infrared-camera.
I was curious about the transmission spectrum of the blue filter, so I measured it with a spectrophotometer. Below is a graph of the measured spectrum. It’s quite the same as on the publiclab website.
The sensor of the Pi NoIR Camera Module itself has a native resolution of 5 megapixels. In terms of still images, the camera is capable of 2592 x 1944 pixel static images, according to the description, but ………… there is more to say.
A chain is no stronger than its weakest link and for the Pi NoIR Camera Module the weakest link is the lens, which has definitely not a 5 Mpx resolution.
The picture below shows the resolution test chart captured using the fluorescent lamp.
For the small lines in the center it is not possible anymore to distinguish the three lines, although they cover 12 pixels.
For infrared the situation is even worse, as can be seen in the image below, which was taken with the infrared floodlight.
The combination of visible light and infrared is slightly better as can be seen below in the image taken with halogen illumination.
