Camera capture using OpenCV

I had a Logitech C120 camera lying around and found that Angstrom had drivers for it. It was about 1:00am and I couldn't wait till I captured an image.

OpenCV is a popular library for image processing. I found this post which got me started with OpenCV. But you must also install opencv-dev to get the header files or the program will not compile.

To get OpenCV:

opkg update
opkg install task-native-sdk cpp gccmakedep
opkg install python-distutils python-compile python-compiler python-devel
opkg install ffmpeg-dev
opkg install opencv
opkg install opencv-dev

My wireless connection was not very good, so I got several package download failures with "wget returned 1". I just switched to USB Ethernet and it went smooth.

Also, if you get "But that file is already provided by package xxx" errors while installing opencv-dev, just uninstall opencv and try again from step 5. I am not sure what the problem was, but it worked the second time.

Finally, around 3:00am I captured the first images. I went to bed both satisfied and eager to get back to my desk and track a ball or something!

Development using Netbeans

Angstrom comes with DropBear SSH a lightweight SSH. But it is recommended to use OpenSSH instead.

To install openssh:

opkg -force-removal-of-dependent-packages remove dropbear
opkg install openssh openssh-sftp-server

Tim Pitman has a neat article that explains how to setup Netbeans for remote development on a MAC.

To share the rootfs on Windows:

1. Configure samba (/etc/samba/smb.conf) to allow access to the rootfs. You can do this by adding the following in the "Share Definitions" section.
   

   [public]
   path = /
   public = yes
   writable = yes
   printable = no

   
   
2. Map \\BEAGLEBOARD\public as a network drive in Windows.

BeagleBoard Networking

Wired networking
The BeagleBoard-xM has an in built Ethernet port. But I used a Linksys USB300M with my C4 version. Angstrom has drivers for the ASIX chip in the USB300M and it installs automatically as eth0 and defaults to DHCP. On Win7, I bridged this with my wireless connection to get internet access for my board.

I assigned a static IP to eth0 by adding this information to /etc/network/interfaces file:

#Bring up the interface automatically. 
#Otherwise, must use "ifup eth0" or "ifconfig eth0 up" everytime.
auto eth0
#iface eth0 inet dhcp
iface eth0 inet static
address 192.168.1.105
netmask 255.255.255.0
gateway 192.168.1.1

Wireless Networking
For WiFi, this table shows the supported chipsets and vendors. I use a Netgear WG111v3 adapter which has the rtl8187 chipset.

auto wlan0
iface wlan0 inet static
wireless-essid xxxxYourESSIDxxx
wireless-key xxxx
wireless-ap xx:xx:xx:xx:xx:xx
address 192.168.1.10
netmask 255.255.255.0
gateway 192.168.1.1

Bringing up the Beagle

One of the important things to have for development on the BeagleBoard is another Linux machine. I have a Windows 7 PC, so I installed an Ubuntu virtual machine using Sun Virtual Box. Using the virtual machine additions, you can share USB devices and folders with the virtual machine.

I managed to get Angstrom up and running on a 2GB SD card following instructions from the beginner's tutorial here.

Once you log into the beagleboard, change the root password using the passwd command. This is a must for SSH-ing into the board.

Plan upgrade

I came across the BeagleBoard a few months ago and I was amazed by its capabilities. It also spurred my interest in Linux development.

So, I decided to use the BeagleBoard to put some raw processing power into my robot. Of the several advantages of this board in a robotic project, I found the following most attractive:

• 720Mhz ARM based SOC(C4 version).
• Only 3"x3" size!
• Can run Ubuntu! (Although, I feel Angstrom is faster given the amount of ram).
• Wireless remote development using an user friendly IDE (I use Netbeans on Windows).
• TI DSP Image processing libraries and all the fun stuff in Linux.

So I ordered one from Digikey and the associated hardware. This guide is a good place to check out while waiting for the board to arrive.

A change of base

I have been busy with my thesis work and other stuff since my last post. There was disappointing discovery a few months ago. The chassis I built earlier couldn't take the weight of two servos, a wireless camera and 11 batteries. The tracks would keep slipping off while turning on carpets :| . Bummer! There went almost a month of searching, waiting for it to arrive and building it.

And so I decided to search for a prebuilt and tested chassis to jump into making things work as soon as possible. I found the RP5 chassis. Its sturdy, comes with 7.2V motors ( = less noise), battery holder and the coolest fact of all - this thing can carry more than its weight, turn on carpets and still keep its tracks on. Plus it comes with these neat accessories to build multiple levels. Here is a pic of it with my boards on.

The chassis

I wanted Takbot to be able to make 360° turns at a spot. This makes programming the movements simple and intuitive, apart from helping to maintain position and determine orientation using a simple wheel encoder. For this I chose a tank-type chassis, mainly cause it would run over things with great fury and awesomeness - if it should decide so. :)

After a lot of digging online, I decided to make my own using the Tamiya track and wheel kit from Superdroid robots. This is cheaper and apparently has more traction than the Tamiya's plastic track set . I used a couple of Tamiya's universal plate sets to make a chassis wide enough to accommodate my huge ATmega128 board (I made this one back when i didn't have the tools for SMD soldering).

Here's what it looks like: