Implementation environment: Ubuntu16.04+Kinetic

ORK function package download and install

This experiment is performed under Ubuntu 16.04. Since this version of kinetic does not integrate all ORK function packages, the following steps are required to install the source code.

1. Install the dependency library

sudo apt-get install meshlab
sudo apt-get install libosmesa6-dev
sudo apt-get install python-pyside.qtcore
sudo apt-get install python-pyside.qtgui

2. Download the rosinstall file to compile

I figured out this step after thinking about it for a long time. It turned out that I downloaded it according to Hu Chunxu's textbook, and I didn't download it for a long time, but the original file could no longer be downloaded, so I went to github and searched it again. The file link is here :
https://github.com/wg-perception/object_recognition_core
To create a new ork_ws workspace, I download ed the ork.rosinstall.kinetic.plus file directly to ork_ws, and then proceeded as follows:

cd ork_ws
wstool init src [file path]/ork.rosinstall.kinetic.plus

3. Download function source code

cd src
wstool update -j8
git clone https://github.com/jbohren/xdot.git
cd ..
rosdep install --from-paths src -i -y

4. Compile

cd ork_ws
catkin_make

It takes a long time to compile...
Then add the environment variable to the .bashrc file

object recognition

The following is the identification process:
(1) Create a model of the object to be recognized
(2) Train the model to generate a recognition model
(3) Use the trained recognition model for object recognition

1. Create a database

Install CouchDB tools:

sudo apt-get install couchdb

Test if the installation was successful:

curl -X GET http://localhost:5984

Create a cola model data:

rosrun object_recognition_core object_add.py -n "coke " -d "A universal can of coke " --commit

You can view it in your browser at the following URL:
http://localhost:5984/_utils/database.html?object_recognition/_design/objects/_view/by_object_name

2. Load the 3D model

Download the existing coke.stl model and download it through github:

git clone https://github.com/wg-perception/ork_tutorials

Download to the src file;
Then load the cola model into the database:

rosrun object_recognition_core mesh_add.py 49cce25ad1745bc78d8c16a90200008e [path]/ork_tutorials/data/coke.stl --commit

View the model:

sudo pip install git+https://github.com/couchapp/couchapp.git
rosrun object_recognition_core push.sh

Click on object_listing:
Click on meshes:
The above is our cola model.

3. Model training

With many models loaded in the database, we need to train to generate matching templates.
The command is as follows:

sudo object_recognition_core training -c `rospack find object_recognition_linemod`/conf/training.ork

4. 3D Object Recognition

This step is unstable and needs to be changed