Detect small smd part

asked 2017-08-21 07:39:09 -0600

I'm trying to develop some vision application to detect presence of small smd component. I already try some approaches, and would like to have opinions from all of you, what is the best one. 1. Template Matching, seems to have lot of fake recognitions, and since the scale and orientation may change it will fail. Right ? 2. HOG 3. Haar cascade seems to me the better, but with lot effort to train the cascade. What should I use inn your opinion?

image description

Comments

hog descriptors or haar cascades will fail with ~~scale &~~ pose variation, too.

berak ( 2017-08-21 07:52:38 -0600 )edit

add a comment

answered 2017-08-21 08:54:10 -0600

StevenPuttemans

20029 ●16 ●82 ●207 http://stevenputtemans...

@berak ah comon that is not true :) the cascades have a multiscale detector. In many cases one has a fixed camera, so the number of scales should be reduced with minSize and maxSize parameters. Then apply a rotation approach to your input image, do detection again and warp the results back. I have suggested this approach for OpenCV 3 Blueprints, right here.

edit flag offensive delete link

Comments

^^ yea, right. (i'm obviously having a dumb day..)

berak ( 2017-08-21 09:04:03 -0600 )edit

Hello Thanks @StevenPuttemans and @berak, for the reply. From what i understand @StevenPuttemans i need to train a haar cascade for the part i want to find, and then just apply on your code. For training could you advice me.

HIT_Braga ( 2017-08-22 05:57:12 -0600 )edit

add a comment

Detect small smd part

Comments

1 answer

Comments

Links

Question Tools

Stats

Related questions

Detect small smd part edit

Comments

1 answer

Comments

Links

Question Tools

Stats

Related questions

Detect small smd part