Geometric Correction
Mathematical transformation compensates for perspective and lens distortion in digital images captured during board inspection. An image warping algorithm translates the pixel coordinates of a distorted raw capture into a corrected, rectilinear grid that matches the original computer-aided design data of the printed circuit board. This adjustment is necessary to prevent false defect calls on components near the outer edges of the field of view where lens distortion is most severe.
The corrected image is then handed off to the analysis software for component alignment and solder joint inspection.
Algorithm Execution
Processors apply the spatial transformation by calculating new coordinate values for every pixel based on a predefined set of polynomial calibration coefficients. The algorithm uses bilinear or bicubic interpolation to estimate the intensity values of the new pixels, which avoids the creation of visual artifacts that could confuse the inspection system. This step is completed in milliseconds by dedicated graphics processing units to keep pace with the high speed of the assembly line.
Because of this rapid processing, the algorithm does not act as a bottleneck in the automated inspection sequence.
Correction Constraint
The effectiveness of the transformation depends on the accuracy of the initial camera calibration. If the calibration coefficients do not match the physical distortion of the lens, the image warping algorithm will introduce geometric errors instead of resolving them. This limitation means the system must be recalibrated whenever the camera or lens is adjusted.
