Small component (around 2mm x 1mm) measurement

Thickness Measurement on friction Stays

One station of the friction stays automation is detecting the thickness and two gaps using CCD with the back light.

The algorithm using the binary cut technique with the edge filter is stable in easily ROI setting parameters.

Left Type:

Right Type:

A gap length detector from the top view of multiple types

To detect the friction stays gap length, some locating and clustering methods are used with simple setting factors.
The effect is shown below:

Left three Types:

Right three Types:

A heavy scratching detector (AUTO ROI with spotlight)

This is a heavy scratching detector.

I implement the detector using the gabor filter and certain binary cut technique with a template.

The gabor filter kernel and the template:

Heavy Scratching:

Light Scratching:

Another hole detector

Detect the focus area by light, find the hole with the low gray limited in the hole boundary range.

A hole detector

To detect the complete screw hole ignoring the overly inside on the disc reader.

 I use the multilevel histogram analyzing method to finish it.

A rotary cylinder auto screw locker

I made the system software -  a rotary cylinder auto screw locker with two air spanner and six-axis machining center (4 for air spanners, 1 for rotating the rotary cylinder and 1 for camera moving.)

The vision algorithm locates the position of screw holes, the system detects the height of the screw by touching the screws which monitoring the differences between the motor pulse and driver command pulse with the air spanner. 

It's fun in monitoring the pulse difference in various velocity and loading though the progress is totally slow.

  

A Rain Removal System

We developed a Rain Removal System with better removal result:









=== Update in 2015/6/7 =====================
Hi every:

For many oversea friends querying me, I provide my research (but written in Chinese) here:
Main Page in NCU Taiwan
PDF File
My code is written in C++ but so ancient and need time to be reorganized.
I think my mainly contribution is on the intensity removal filter in rain area;
But some learning patterns need to be improved.

Hope this is helpful on you!
=======================================