Pick and Place Machine Vision Calibration for Fine Pitch Components
Fine pitch machine vision calibration requires quartz grid targets, telecentric optics, and closed-loop thermal drift compensation to hold 5 µm accuracy.
Brown-Conrady Model
Optic Distortion
Geometric correction math adjusts the spatial mapping of images to remove radial displacement caused by lens geometry. The brown-conrady model provides a mathematical framework to represent these radial and tangential errors by expressing the displacement of image pixels from their ideal locations. This formulation assumes a perfect pinhole projection and calculates polynomial offsets to shift pixels into their correct relative positions.
Correction Parameters
Coefficients within the brown-conrady model quantify the degree of pincushion or barrel distortion present in a specific optical system. Engineers solve for these variables by imaging a grid target and identifying the deviation between captured lines and straight reference vectors. Higher order terms account for asymmetric tangential effects where the lens axis fails to align perfectly with the sensor plane.
Standard algorithms employ iterative optimization to minimize the residual error across the entire field of view until the output image matches a rectilinear projection.
System Application
Automated inspection equipment relies on these calculated values to ensure geometric accuracy during high speed board component placement. Precise registration between the camera sensor and the mounting head allows the software to subtract distortion from every frame before the vision processor executes a solder joint check. Consistent calibration routines check for lens heating effects that shift these values during long shifts on the assembly line.
Correcting pixel geometry before feature extraction prevents measurement bias in fiducial alignment or pitch verification. Any failure to maintain valid coefficients during production leads to coordinate offsets that degrade the placement quality of small surface mount devices.