This week we have looked at how we’re going to achieve object detection, so that car will be able to recognize its target, both initially before it is thrown and when it goes looking for it. To perform this task we’ll be using the ROS package find_object_2d (which also has a 3d mode, when using a stereoscopic camera setup) which is a ROS implementation of the Find-Object Project, which is an OpenCV implementation of the feature detection algorithms SIFT, SURF, FAST and BRIEF. We tried these keypoint detection algorithm to find which one would perform better on our system, we have not gotten our cameras setup on our system yet so we will not be able to decide which algorithm we are going to use, and we’re not able to test the 3d mode yet either. Currently we are leaning toward SURF due to its increased efficiency in terms of computing power. 

Sadly the cameras we ordered was lost in shipping and had to be reordered so that will take some time before we can try the system out live with two cameras. Until then we will have to use pre recorded bags that is provided on the ROS wiki page. Other than the cameras all hardware is now completed and we only have software to go before we have a working product.