Raspberry Pi Camera

Pick-and-Place Vision

The 5MP resolution of the Raspberry Pi Camera provides a wealth of information for pick-and-place (PnP) operation. Although humans love video, we will rely on still pictures for pick-and-place. Still pictures provide greater control over the entire imaging process. There is also no need for the motion-tracking capability of video. For this reason we will use:

raspistill

Prerequisites

Before proceeding, you will need:

• Raspberry Pi 2 Model B. The original Pi is workable but not recommended.
• Raspbian Jessie
• A Raspberry Pi camera. The Pi Noir is ideal.
• A long flat cable
• A good case
• Neopixel 16-LED ring for illumination and FireStep status display (optional)

RAM Disk

We will be processing a lot of images and we want to keep them off the SD card. Lets create a RAM disk with at least 64MB of storage for our images:

sudo mkdir /img
sudo mount -t tmpfs -o 64m tmpfs /img

Parameters

The raspistill program is versatile and we will launch it in continuous, triggered image capture mode using the following command line:

raspistill -w 400 -h 400 -ex snow -awb fluorescent -ev 12 -t 0 -s -o /img/camauto.jpg &

Parameter	Description
-w 400	Image width ensures small file size for fast OpenCV processing
-h 400	Image height ensures small file size for fast OpenCV processing
-ex snow	Exposure for a predominantly white background (e.g., paper)
-awb ...	Fixed white balance for capturing image series (auto is too variable)
-ev 12	Increase exposure for brighter image over auto exposure
-t 0	Continuous picture capture
-s	Wait for SIGUSR1 to take a picture
-o ...	RAM disk image output file (i.e., /img/camauto.jpg)

Once raspistill is running in the background, we can take a picture using the following simple bash script that sends a signal and waits 300ms for capturing the image, which it renames as /img/img01.jpg:

PID=`ps -C raspistill -o pid=`
kill -SIGUSR1 $PID
sleep 0.3
mv /img/camauto.jpg /img/img01.jpg

The 300ms capture interval accounts for the shutter speed (1-100ms) as well as the JPG compression time. Faster times can be achieved by specifying -ISO 800 and using the Pi camera as a light meter (-set) to determine shutter speed (-ss) for a series of captures. However, 300ms is certainly fast enough for many applications. The method described here is simple to use and works for different camera heights from the bed.