OpenCV library is used by other applications for image enhancement, panorama stitching, object detection and recognition, etc. OpenCV Manager provides the best version of the OpenCV for your hardware. It also receives the latest stability and performance updates for the library.
This app uses the BackgroundSubtractorMOG algorithm from OpenCV's android library to identify objects in motion from the camera and draws a red overlay around/over them. The ideal is to have nice contours drawn around the objects that are moving. A slider is also added for the user to adjust the learning rate, which may or may not have any effect on the contours.
In order to utilize the app, the latest version of OpenCV Manager (also found on Google Play) must be downloaded and installed correctly. This app has been successfully tested on Android 2.3.4 and Android 4.3.
At the moment this can considered an initial tech demo, please feel free to play around with it. The source code for this implementation can also be found in github link below!
This app compares the reference image to the images in the added list. It uses ORB feature detector with BRIEF descriptor extractors. It uses BRUTEFORCE_HAMMING method to match the descriptors. Also, based on preference, app can also check the homography of the matched keypoints with LMED (least of median squares) and remove the outliners to give a better result.
How to use:
1 - Move the vision wherever you want, take pics along the way by using ADD button.
2 - You can also take a reference picture, that you want to be recognized. When you click it, it will be shown on the right of the camera view.
3 - When you are done, and make sure you have added images and have a reference image, click the Find Match button. It will calculate the keypoints of each image and the reference image and apply the matching algorithm.
4 - After matching and comparing, it will show the best match among all matches.
5 - You can also use Homography Off/On switch to provide more accurate results.
6 - Also, you can use Images only/with matches switch to show which keypoints are matched. REMEMBER, you have to click find match again to see the effect of Homography and Matches switch.
This app is still in development, actually being used by another project but I put it on for cool demonstration!
Source code available: https://github.com/mustafaakin/image-matcher
Sourcecode available on https://github.com/iisy/SmartObjectRecognition.git
Code is the rewrite of https://github.com/MasteringOpenCV/code/tree/master/Chapter2_iPhoneAR using the "OpenCV for Unity".
・Texture2DMarkerBasedARSample - By detecting the marker from Texture2D, display AR model.
・WebCamTextureMarkerBasedARSample - By detecting a marker from WebCamTexture, display AR model in real time.
Project Website: http://boofcv.org
For instructions and a more detailed explanation:
Full source code:
The Enabler takes care of a lot of boilerplate code, so your app's code can be simpler. It also includes a basic GUI for most functionality of the Android library.
The Enabler also has views that allow you to:
- Display all received simple vehicle messages from the VI.
- Display all received low-level CAN messages from the VI, if that functionality is enabled in the firmware.
- Send an arbitrary CAN message to the vehicle through the VI (requires firmware that supports this feature).
- Send a diagnostic request to the vehicle through the VI.
More information on getting started with OpenXC on Android is available at http://openxcplatform.com
This app sends crash reporting data to Bugsnag for the purposes of assisting the OpenXC maintainers with debugging. If you do not wish to have your crashes reported, please compile the app from the sources.
The source code for the OpenXC Enabler is available at https://github.com/openxc/openxc-android and is provided under the BSD license.
This app was used for augmented reality time, distance and other tests and reserches in DonNTU, Donetsk.
This libraries are used in project:
Load scaled pictures
Color plane extraction - RGB, HSV, HSL, XYZ, YCrCb
Binary convolution functions like Dilate, Erode, etc
Save post-processed pictures
Simple edge detection should be fluent for everybody but face detection need more processing power.
Use Neon for colorspace transformation
Face detection optimization is done thanks to Bill Mc Cord