I am currently working on a Project to detect a Chessboard and also the position of the figures. After taking a photo, the App should scan for known figures (pawn etc.).
My Question: I found two methods of detecting known Objects in an image.
1. Extract features from one image and search for same features in a second image (Features2D + Homography to find a known object)
2. Search for known Objects using Haar cascades (Cascade Classifier Training)
I don't know which of these methods is more reliable. Can I use the Homography to detect multiple objects in one image?
Related
I am working on a project where I have to recognize objects in a grocery shells. You can see the sample image below:
I need to find what products exists in an image. The example of result image is shown below:
OpenCV tools like SURF, SIFT, ORB detects only one occurrence of the object in an image. Can you suggest some papers or tools to solve this problem.
Normally there are multiple techniques to detect multiple instances of the same object in an image.
The most primitive way to do that is template matching. So you create a database of training images at multiple scales and rotations to be able to detect such objects in varying conditions. But there are many techniques that are better than such legacy technique.
Some other techniques uses texture features that is invariant over scale, rotation, or both. For example, GLCM, LBP, HOG, SIFT, ORB and others.
Your statement OpenCV tools like SURF, SIFT, ORB detects only one occurrence of the object in an image. needs more enhancement.
The listed tools are not intended to detect objects but they are means to extract texture features.
You are the one to adjust them to detect multiple objects.
There is a more fine way to solve your problem. It seems that all of your objects that are required to be detected contains the text TASSAY.
you can easily extract that text using a group of morphological operations and then using a blob detector detect the location of the text.
After returning the text, it can be easily to measure the text location.
The object bounding box can be easily inferred from the text location.
Hope it helps.
Im looking at integrating an AR Kit into our iOS App so we can use the camera to scan a room or field of view for objects. Above is an example of what i mean, if you were to bring up the camera it would highlight the separate objects in the room and allow them to be clicked and "added" into the system.
Does anyone know if this is achievable with the current AR kits or anything else out there? It all seems to be the fact that objects that you are looking for have to be pre-defined and loaded into a database so the app can find them. Im hoping it should be able to pick out the objects realtime. It doesnt need to actually know any details on the actual object just so that can be pulled off the base scenary.
Any ideas?
OpenCV library (iOS) contains many algorithms to compare different image blobs. If you want to match some simple template to find objects then try Viola & Jones algorithm and so called Haar cascades. OpenCV has trained collection of templates in XML files for detecting faces for example. OpenCV contains utility for training thus you are able to generate cascades for other kinds of objects.
Some example projects:
https://github.com/alexmac/alcexamples/blob/master/OpenCV-2.4.2/doc/user_guide/ug_traincascade.rst Cascade Classifier Training
https://github.com/lukagabric/iOS-OpenCV Example code for detecting Colors and Circle shapes
https://github.com/BloodAxe/OpenCV-Tutorial Feature Detection (SURF, ORB, FREAK)
https://github.com/foundry/OpenCVSquaresSL Square Detection using Pyramid scaling, Canny, contours, contour simpification
My question is - can I recognize different templates in a source image using feature detection in OpenCV? Let's say my templates are road signs.
I am using ORB, but this is not tracker-specific question.
My basic approach without feature detection is:
Image preparation (filtering etc);
Detecting ROI where my object may be located;
Resizing ROI to templates' size and comparing with each template I have (ie. template matching);
Maximum correlation after comparison is an object I look for.
But with feature detection I detect keypoints and descriptors for each image in my template set and for my ROI where object might be located, but matcher returns distances for all descriptors I have in my ROI.
I can't tie this to any correlation between ROI and templates, or, in other words, I can't decide whether ROI image and template image are the same objects based on information provided by matcher.
So, to be more specific - is my approach wrong and feature detectors are used to detect one template object in a source image (which is not what I need) or I'm just not grasping the basic concepts of feature detection and thus am in need of help.
You may be missing two aspects. One is to remove outliers in your feature matching using a method like RANSAC+homography. The second point is to project the corners of your template to the scene, to make a "rectangle" of your image. Also you should define a threshold on how many inliers you will consider the minimum for a right detection.
Check this tutorial on finding objects with feature detection.
I will refer you to a book called:
'opencv2 computer vision application programming cookbook'
Just browse the relevant chapters.
I am trying to do object recognition using algorithms such as SURF, FERN, FREAK in opencv 2.4.2.
I am using the programs from opencv samples without modifications - find_obj.cpp, find_obj_ferns.cpp, freak_demo.cpp
I tried changing the parameters for the algorithms which didn't help.
I have my training images, test images and the result of FREAK recognition here
As you can see the result is pretty bad.
No feature descriptors is detected for one of the training image - image here
Feature descriptors are detected outside the object boundary for the other - image here
I have a few questions:
Why does these algorithms work with grayscale images ? It is apparent that for my above training images, the object can be detected easily if RGB is included. Is there any technique that takes this into account.
Is there any other way to improve performance. I tried fiddling with feature parameters which didn't work well.
First thing i observed in your image is, object is plane and no texture differences are there...I mean all the feature detectors you used are for finding corners which are view invariant, it means those are the keypoints in an image which are having unique neighborhood and good magnitude of x and y derivatives. I have uploaded my analysis...see the figures
How to know what I am saying is correct?
Just go to the descriptor values of a keypoint you find over your object and see the values, you will see most of them are zeros...Because a descriptor is the description of variation of the edges around a corner point in a specific direction (see surf documentation for more details).
The object you are trying to detect is looking like a mobile phone, so you just reverse the object or mobile and repeat the experiment and you will surely get good results...Because on front side generally objects have more texture like switches, logos etc..
Here is a result I uploaded,
I have used Haar classifier with OpenCV before succesfully. Unfortunately it seems to work only on square objects and fixed angles (i.e. faces). However I need to find "long" (rectangular) objects which have different angles (see sample input image).
Is there a way to train Haar classifier to find such objects? All I can find are tutorials for face recognition. Any other alternative approches?
Haar classifiers are known to work with rigid object only. You need a classifier for each of the view. For example, the side-face classifier in OpenCV doesn't work as good as front-face classifer(due to the reason being, side face has more variation in yaw-pitch-roll than front face).
There is no perfect way of answering your question.
However, in your case whatever you are trying to classify (microbes I suppose) are overlapping on each other. Its a complex issue. But, you can isolate the region where microbes occur (not isolate each microbe like a face).
You can refer fingerprint segmentation techniques that are known to enhance the ridges on a fingerprint (here in your case its microbe edges) from the background and isolate the image.
Check "ridgesegmentation.m" in the following page:
http://www.csse.uwa.edu.au/~pk/Research/MatlabFns/index.html