Large and expensive specific hardware are usually required for tracking systems based on Computer Vision. This circumstance limits their extensive application in industrial processes. With the aim of overcome this fact a low cost image processor has been implemented. This specific one board processor permits a standard industrial PC to perform the tracking of some relevant local features at video rate (50Hz). Local features in the image are transformed to a polar representation in order to simplify the recognition and location processes in front of rotations. This transformation is provided in real time by the presented image processor so the host only has to apply the recognition algorithm. For image processor implementation FPGA technology has been used.