Geometric pattern matching issues

[Sonofsam]

Hello there,

I have two basic questions, and I was hoping if someone could push me in right direction. 

1) I am trying to do pattern recognition(edge, and golden template training) on my car's dashboard panel, but the problem is that the dashboard icons or warning lights are small, and camera is at a distance. so I have to keep the edge threshold very low in order to detect their presence(or absence), but keeping the edge threshold low (i.e 2 - 5) makes the detection process very sensitive, and unstable at times.  Since it's a binary inspection (icon is lit up or not) I was wondering if there was some stable way to detect their presence ? is there any setting(filter etc) i should be aware of that can make the process more stable? 

2) The bounding box/rectangular frame is changing sizes and disappearing in no time, it never gets stable. can any one help me with that.

attached are the code for bounding box and unpredictable frame result i'm getting.

Thanks.

[The_Hunt]

Hello SonofSam, 

I am assuming that you are unable to move your camera closer to the target in order to achieve a higher resolution? 

Since this distance is creating issues with the stability of your edge detection code, perhaps you may have better results if you try and identify the boolean status of the lights by using color pattern matching instead. 

There is a great shipping example with the Vision Development Module that you could use to test this out. 

Help >> Find Examples >> Toolkits and Modules >> Vision >> Functions >> Color Pattern Matching Example.vi