Feature-based reverse engineering

technical reportReverse engineering entails building a representation of an engineered object using only information derived from an instance of that object. This research has focused on the reverse engineering of physical objects, specifically machined parts. Reverse engineering differs from many other sensing problems in that the results need to be very accurate (to within as little as one thousandth of an inch) and describe the entire surface of an object. Speed and low cost are also desirable attributes for a reverse engineering system. Many researchers and engineers focus on free-form surfaces and use sophisticated representations for their shapes. Machined parts, however, are often composed of relatively simple features, such as holes and pockets. The combination of these simple shapes can produce objects which are difficult to reconstruct, particularly with representations that are general enough to handle free-form surfaces. Rather than using a "one size fits all" representation and a complex autonomous system, this research takes a more direct approach, essentially creating a set of "electronic calipers" to be used as a smart measuring tool. This allows use of different representations for different portions of an object's surface. From the feature-based CAD community comes a set of representations that is aptly suited to the class of objects used in this research. By using these representations, the output from the system can be machined with minimal human intervention. This system is expected to be robust in the presence of noise and outliers that come with the use of low-cost sensors. Additionally, it could be used as a platform for driving more accurate sensors, such as a CMM. The same feature-first strategy used here could be used to get touch sensing accuracy with a minimal number of touches