Vision Assistant - Object overlap operation

[Kriton]

Vision Assistant performs logical and mathematical operations which can compare two images on a pixel by pixel basis.  I was wondering if it has, or could readily be improved to include, object based operations?  (The object referred to here is a 2D region in an image with some distinctive attribute, e.g. a particle, rather than a term for a software datatype.)  I am particularly interested in an overlap function, whereby objects of one image that overlap objects in a second image are kept.  Potential applications (all of which I am interested in) of this type of operation include video tracking of moving objects (particularly ones that can change shape), particle filtering by danielsson, and a spatially defined method of object noise filtration.   

[VuD]

Hi Kriton,

Besides the lower level pixel comparisons, Vision Assistant offers other image processing functions such as pattern matching and particle detection.  When using pattern matching, you create a template that represents the object for which you are searching. Your machine vision application then searches for instances of the template in each acquired image, calculating a score for each match. This score relates how closely the template resembles the located matches. Pattern matching finds template matches regardless of lighting variation, blur, noise, and geometric transformations such as shifting, rotation, or scaling of the template. 

You can use particle analysis to detect connected regions or groupings of pixels in an image and then make selected measurements of those regions. You can extract particles from a grayscale image by thresholding the image into background and foreground states. Zero valued pixels are in the background state, and all
nonzero valued pixels are in the foreground. Particle analysis consists of a series of processing operations and analysis functions that produce information about particles in an image. Using particle analysis, you can detect and analyze any 2D shape in an image.

You can find more information about these and many other image processing functions (including algorithms/logic used) in the Vision Concepts Manual, which you can access by going to Start >> Programs >> National Instruments >> Vision >> Documentation >> Search NI Vision Bookshelf >> Vision Concepts Manual.

I hope this helps give you a good start. 

Vu

[Kriton]

Thank you Vu.  I do appreciate your time and thought.  I am using VA already, and have read through the concept manual.  However, my question was along the lines of improving/expanding the capability of VA as a tool.  I am a scientist, not a programmer, so I understand the image analysis concepts, and even algorithm development, but not how to make them happen in C++.  I would prefer to continue using it as a tool, but expanded to be a little more flexible.  Currently, it is quite flexible when it comes to pixel analysis.  I want greater flexibility with object analysis.

I'll give an example.  Take a video of two bouncing balls.  You can color-threshold the image to identify the balls.  If you look at sequential images, and if the velocity isn't too great, by identifying the overlapping objects in each image pair, you can track both balls through a long sequence of images. 

Now, it may be possible to do this by an OR of the two (threshold) images, followed by an AND between the result and the second image.  However, when in a noisey environment, this can add a lot of objects that are irrelevent.  It would be a lot simpler to say "If the objects in the first image overlap the objects in the second image, then keep these objects, otherwise cut them out."  This is effectively an AND in object space. 

So by having logical operators in object space, OR, AND, XOR etc. would improve the flexibility of this tool.  So my question is perhaps twofold - does NI consider improvements from customers?  And, have they (or will they) considered this improvement?

Message Edited by Kriton on 07-19-2007 09:03 PM

[VuD]

Hello Kriton,

I apologize for overlooking your question regarding customer feedback.  The R&D at National Instruments takes user suggestions into strong consideration as they strive to improve and develop our current and future products.  You can directly submit a product suggestion from our website at htttp://www.ni.com/contact.  There is a link for product feedback which I have included here.