Lately I have been using numpy way too much and there’s no point in doing things non-pythonic way so I always look for some cool tricks using which I can increase the speed and get the correct output. I require numpy basically for Image Processing and Computer Vision. I always forget these tips and have to look up again on stackoverflow, so I am compiling a list of all such tricks that I have used.
»I recently got a Kinect to work on at my sumemr internship at Ducere Technologies. Having heard so much about OpenNI, I tried installing it on my Ubuntu 12.04 LTS 64 bit machine. It took some time to configure, build and install it. I got help from few blogs. I managed to install it somehow. Connected the kinect and fired up the demo program and ran into troubles. I was getting following error.
»I had proposed a project for SimpleCV in this year’s Google Summer of Code and it has been accepted. This year my project is concentrated on improving SimpleCV’s core functionalities and adding support for shape matching and detection which is one of the most important purposes of Machine Vision in industrial automation.
»
So I’m currently working on HTC Evo V 4G and was desparately trying to obtain images from both the camera. One thing was sure that I couldn’t use OpenCV’s Java Camera or Native Camera (it doesn’t even work with ICS). I decided to use Android SDK Camera. I tried posting question on stackoverflow and OpenCV forum, but couldn’t find any proper solutions. I tried taking pieces of code from wherever I could and wrote something, but it wouldn’t work. I also found a perfectly working code but it was giving me Static Linkage Errors. It meant the OpenCV manager couldn’t be loaded in the application. I had done everything step by step but it wasn’t working.
»Whenever working with stereoscopy, it is a necessity to calibrate the cameras and get the required intrinsic and extrinsic parameters.
»As I had mentioned in earlier posts that I was working on Stereo Images, disparity and depth images, I will elaborate about disparity maps and show how to compute it using OpenCV.
»Many people who are trying to search for Android Opencv stumble upon my blog due to the name Paranoid Android. I had decided that I would never make an android appliaction as I hate Java, but desperate times require desparate measures, and so I installed Android OpenCV and created a small application.
»Recently, I have been working on Stereoscopy, disparity and depth maps, and my primary objective being getting depth, height of the object. In a nutshell, I am working on obtaining 3 dimensional data from the scene. As I was a bit swamped with all the work, I didn’t get time to write about it. I will describe and elaborate more on stereoscopy in upcoming posts.
»In the previous post we discussed Hough Transform and how to implement it to find lines. To get better accuracy we need to compromise on the computing front. It takes a lot of computation power to iterate over all the points and add vote. To reduce this computation, researchers came up with some probabilistic techniques which would increase the computing speed without losing much accuracy.
»Let’s say you take a snapshot of a window. You used Canny/Soble edge detector and found a nice boundary of the window and you want to figure out representation of the window’s edge, size, etc. How would you do it? Right now it’s just some 200 odd pixels in near vicinity and representing the shape of a line.
»What is Open Source To Me?
»