Real-time automatic target location by template matching on blurred images

High precision manufacturing is dependent on the accurate positioning of the part prior to each processing step. The part is typically held in a transfer fixture which many be used either simply as a part holder or as an alignment reference. A typical application is explored in this work, grinding of slots in a thick ceramic wafer mounted on a part fixture. A 1 micron slot is located with a position tolerance of 10 microns in the nickel-zinc-ferrite wafer. Part location is determined by means of a 50.8 by 1 micron target deposited onto the ceramic during an earlier process. In the current process, the target is magnified by a 250X vision system and the machine operator manually aligns the part using a set of cross hairs overlaid on the image. A high precision x-y stage is used to adjust the part location. However, the use of a fixed focal distance camera combined with part mounting variation results in an unfocused image for the operator. Manual alignment of the unfocused image leads to target location error during centering of the target in the cross hairs. The misalignment, in turn, results in process variation. The location of the slots varies based on alignment and reduces the process capability. To reduce process variation, an automated target location algorithm has been applied. The algorithm uses template matching to detect the target. This simple algorithm has been shown to locate the target within the required tolerance in spite of image blur. Using low cost processing, the system is able to determine the target location in real-time. For real-time control, the algorithm must determine the x-y coordinate of the target in 10 seconds or less. This effort shows the potential for a simple location algorithm to be implemented in a manner which can significantly decrease process variation in a precision fabrication process.