Calibration method for camera

Your machine will have to have a reference position. From this reference position, you determine all other machine positions. If the camera is attached to the head, you could place a glass slug and measure camera-nozzle offset, by placing the slug with the head nozzle and measure the X and Y distance with the camera. You can also lower the nozzle in silly putty and leave an inprint of the nozzle. Move the camera over this inprint and measure the offset with the encoder of the X and Y axis.

reply »

My favorite kind of calibration is where the machine learns what it did incorrectly. Then fixes itself!

However, there are many types of machines that rely on human feedback as to the results of what it did.

nozzle calibrations involve many factors. are you using CCD or line-scan camera's? Are the image chips analog or digital? This will determine how complex your system will become....

1. vision height vs. mounting height. If the vision height is different than the mounting height, your camera must be perpendicular to the nozzle (z-axis). If not, then an x-y deviation will result during mounting. However, a correction table could be used to eliminate the camera alignment OR even simpler is to have the vision height EQUAL to the mount height. By putting both operations in the same z-axis position, you eliminate the x-y shift as a function of z-height.

2. upwared looking camera position is adjusted or "bore-sighted" as a function of mounting a precision glass slug. The error of the mounted slug is used to determine the offset postion. This part is mounted at 4 major rotation angles to determine tool run-out (eliptical tool path). Eliptical tool path x-y deviations can be eliminated by always visioning the part AT the final mounting angle. This may be time consuming or not possible if this is a multi-head system using a line-scan architechure.

There are many techniques for sighting a camera. Some are more accurate than others. What is your target accuracy / repeatability?

reply »