Thursday, August 25, 2016

RPLiDAR + ROS + Hectorslam Setup

So, I have been lately working on a very interesting but difficult project with team UAV Nirma , 'UAV navigation in Indoors' and the timeline is till August 2017. 

The first thing you might be asking is why such a costly and complicated setup?
The thing is, it is not as complicated as you think. The specific algorithm through which we want to devise is the SLAM algorithm.

What SLAM does is that it maps the environment, extracts the landmarks and gets to know where it is using these landmarks. There is a very good documentation available here.

Since the quad has to know where it is with respect to the landmark, we have to use a lidar and to use the lidar - ROS.

The lidar we're using is the RPLIDAR from RoboPeak. 


For the ones who are looking onto how to interface the RPlidar with ROS, here's a very good tutorial.

Here's a video of the hectorslamming:


Monday, August 1, 2016

People Counter using OpenCV and Python

People counter system is counting system based on Image processing techniques to count human traffic at various places like Retail shops, Malls, Public places etc. The module was made using an RTSP stream from an overhead CCTV camera and processed through OpenCV library for Python. This Module was made at my internship.


 



fig.1 Snap of the people counter GUI



 The live video feed is taken frame by frame to process and morphological operations were applied to each frame to get rid of noises and get proper blobs. Next, using contours method we extract the co-ordinates of each blob and track them so that the blob can be counted in a region of frame. The accuracy of the counting depends on a number of factor such as height of camera, field of vision, camera angle, lighting etc..

Take a look at the video: