In this application, a machine vision system has been set up to automatically detect defects and damage to industrial nozzles utilised in the Surface Mount Technology (SMT) industry. Such nozzles are used to pick and place SMT components in the manufacturing of printed circuit boards (PCB).
The high speed and frequency with which these nozzles are used means that they have a relatively short life span and also that they have to be periodically checked for general wear (eg abrasion), defects or other damages. In this situation, an automated machine vision system is well suited to perform quality checks, indicating which nozzles have to be rejected.
To identify defects as well as wear, a number of image processing algorithms have been performed. In this particular case, features that were searched on the acquired images included possible shape deformation, edge detection (for instance, in the case of material chipping or erosion) and the presence of clutter (such as 'dirt' found in the nozzles' cavities). In order to take into account varying light conditions and noise in the acquisition of the images, a Gaussian filtering algorithm has been applied.
To extract the bottom face of the nozzle's tip (which is the only area we are interested in) from the rest of the image, an adaptive smoothing algorithm has been used.
On the right, you can see the output of the processed images for two different types of SMT nozzles.