Introduction to Reverse Engineering
Reverse engineering can be defined as the process of creating a CAD model of an existing product or component by capturing the components physical dimensions and surface features through various engineering approaches. Reverse engineering is usually undertaken in order to redesign the product for better maintainability or to reproduce a copy of the product without access to original design from which that product was made.
The main goal of reverse engineering a product or component is to successfully generate a 3D CAD model of that component which can be used for future modeling of parts when there is no CAD model available as well as to generate a clean and smooth 3D model free from noise and other imperfections. Reverse engineering is accomplished in three principle steps shown below;
The first step in creating a 3D model is Data capture or Data acquisition. Data capture is the process of acquiring point coordinates from part surface which results in a cloud of data points stored as an image. It can be done through wide range of available contact or non contact techniques. Data segmentation process is performed in order to extract surface features of the part from already obtained clouds of data points. The point clouds produced by 3D scanners and 3D imaging can be used directly for measurement and visualization in the architecture and construction world. The output of segmentation process consists of labeled points belonging to a particular region on part surface. Holes are filled and noise is filtered out by applying surface modeling techniques.
Case Study- Water pump
Water pump was selected for this Reverse engineering exercise because of its complexity in shape and size measurements as shown below. Water pump has no predefined center of gravity because of its non-symmetrical shape. It means that if the pump is placed on the opposite side it will not be balanced.
Laser Triangulation which is non contact technique for Computer aided reverse engineering (CARE) is the most suitable for this case which will be discussed in detail in later sections. The least appropriate RE technique in our case is the traditional contact based approach i.e. Coordinate Measuring Machine (CMM) which is used for data acquisition because it cannot capture complex part geometry features and dimensions as the water pump is non symmetrical.
Following are the difficulties of RE to create chosen design i.e. Water pump in our case.
• Multiple views
• Fixture (placement)
• Noise and incomplete data
• Surface finish
Laser triangulation is a method which uses location and angles between light sources and photo sensing devices to figure out position. Triangulation range finders have a limited range of some meters, but their accuracy is relatively high. The accuracy of triangulation range finders is on the order of tens of micrometers.
Laser scanning describes the method to scan a surface using laser technology. Several areas of application exist that mainly differ in the power of the lasers that are used, and in the results of the scanning process. Low laser power is used when the scanned surface doesn’t have to be influenced, e.g. when it only has to be digitized. 3D laser scanning methods are used to get information about the scanned surface.
A source light is focused and projected at a pre-specified angle at the surface of interest. A photo sensitive device usually a camera senses the reflection of the surface and then by using geometric triangulation from the known angle and distances, the position of surface points relative to the reference plane can be calculated. The light source (Laser) and the camera are both mounted on a traveling platform in order to get multiple scans of the surface.
Data segmentation process is then performed in order to extract surface features of the part from already obtained clouds of data points. Then surface data is examined by applying surface modeling techniques in order to fill any holes present and noise is filtered out. Finally multiple scans are merged together to get final 3D model of the component.
The process of reconstructing 3D model of the part consists of following steps;
1. Acquiring range images of the water pump
2. Pre processing acquired data (Data Segmentation)
3. Data Post processing also called Data integration
4. Final 3D model
In acquiring range images, CCD camera in IVP range scanner captures the scenes associated with part. As a result a grey scale image is obtained which shows the intersection between the laser plane and the object, i.e. a line is obtained. This process of image acquisition yields number of different images of the object from different viewpoints set by the user. The main purpose of getting multiple images is to eliminate missing data due to occlusion from the water pump. Although this system produces 2D views of the part, they are still considered as single views because one view of an object cannot complete the reconstruction of object.
Pre processing acquired data
After this step, pre processing of the range images is done which means cleaning the data from noise and occlusions.
In the next step, multiple images of the object from different views are obtained. The object is reoriented in front of the scanner to obtain these images. Increasing number of scans results in taking longer time to reconstruct an object, therefore this step must be carefully planned depending on the size and material that make up the object. The individual images obtained must be aligned or registered in a single coordinate system so that they can be easily integrated into final 3D model of the object. This is achieved by accurately tracking scanner’s position and orientation using high accuracy sensors.
The main purpose of collecting multiple views range images is to combine them and register them in order to obtain an entire surface of the object and thus produce a analogous set of parametric surface patches. When constructing surfaces, we want to have overlapping views of the object. These different overlapping views results in omission of occlusions.
Special software which uses ICP (Iterative Closet Point) Algorithm is used to match similar features and points on the different range images obtained. After all the views are obtained, they are post processed for surface smoothing and multiple views registration.
Final 3D model
Finally all these views are merged together at the ideal feature locations. Two methods which are used in the post processing stage are Surface-based Methods and Volumetric based methods. These methods are used for removing redundant overlapping regions. After this step, 3D model is ready for reconstruction of an object.
Results and Conclusion
The three primary steps in Reverse engineering are Data Collection, Data segmentation and Part modeling. Data collection is done using automated devices while segmentation and part modeling is done manually. The operator defines the type of surface to be fit to each data segment. Finally the software is used for fitting the surface. Using laser based triangulation system has many different benefits that make it ideal for 3D reconstruction. A laser based triangulation system is not affected by ambient lighting during data acquisition. Results obtained by laser triangulation are highly accurate, less noisy and have smoother surface textures.
 R. Bardell, V. Balendran, and K. Sivayoganathan, “Accuracy Analysis of 3D Data Collection and Free-Form Modeling Methods”, Journal of Materials Processing Technology,133, pp. 26-33, 2003
 J. Clark, “Implementing Non-Contact Digitization Techniques within the Mechanical Design Process”, Sensor review, Vol. 20, No. 3, pp. 195-201, 2000
 J. Chow, T. Xu, S. M. Lee and K. Kengskool, “Development of an Integrated Laser-Based Reverse Engineering and Machining System”, International Journal of Advance Manufacturing Technology, Vol. 19, pp. 186-191, 2000
 William B. Thompson, Jonathan C. Owen, and H. James de St. Germain, ”Feature-Based Reverse Engineering of Mechanical Parts” November 6, 1995
 Y. Chen and G. Medioni, “Integrating multiple range images using triangulation,” in Proc. ARPA Image Understanding Workshop, pp. 951–958, April 1993