This looks more like RGB channel misregistration than chromatic aberration. In line-scan or unusual CCD systems, the red, green, and blue samples may come from physically offset sensor rows or from a readout pipeline that places the channels slightly out of alignment. That creates colored fringes on sharp edges.

A practical fix is to split the image into separate R, G, and B channels, then shift the channels relative to each other until edges line up. Based on the community answer, the blue channel appears to be the reference, with green shifted by about 5 pixels and red by about 10 pixels vertically, though the exact offsets should be measured for your device.

For best results, calibrate using a test target and determine the true per-channel offset. If the offset is stable, correction is simple and fast: apply a fixed vertical translation to each channel before recombining. That should be feasible in real time on kiosk-class hardware.

So the likely cause is sensor/channel alignment, and the main remedy is per-channel registration rather than optical correction.