عنوان مقاله [English]
One of the most commonly used techniques in obtaining three dimensional measurements is the triangulation technique using structured light, which is based on a simple mathematical algorithm. The current paper deals with 3D scanning and model reconstruction of mechanical parts being produced in a continuous manner. The objective of the current research is to firstly scan and reconstruct the part model and secondly to improve the accuracy of the model reconstruction. A simple test rig is designed and built to scan the object. The test rig is made of optoelectrical and mechnical parts. The optoelectrical parts include a digital CCD camera and a laser light projector. A conveyor belt with controllable constant speed is used to generate the movement required when
scanning. The camera is located above the conveyor belt in a fixed unknown position. The laser light projector is installed on a five degrees of freedom robot arm facilitating the adjustment of the projector to the desired position and orientation.
The final accuracy of the reconstructed model is considerbly affected by several factors, namely; the accuracy of the intrinsic and extrinsic parameters of the camera, the light projection angle, and the uniformity of light
distribution over the entire scene, both in terms of location and time. Camera calibration is performed to identify camera parameters. Twelve parameters were identified in total. The parameters found were verified through measuring previously known 2D objects. Novel simple mirror arrangements are employed to accurately measure the light projection angle.
A triangulation technique, along with a laser sheet of light with Gaussian distribution, is used to generate depth information. The intersection of the sheet of light and the object produces a 3D curve. Deviations of the curve from a striaght line represent the depth information, which can be extracted and translated into height variations using the triangulation technique. The CCD camera is used to capture both the texture and the depth carrying signal. The captured data are transferred, pre-processed, processed and converted to CAD data format. Curve approximation techniques are used to smooth the curve. Consequently, the CAD model is generated in a CAD environment. The model is then exported to an analysis environment for further analysis. To improve the
accuracy of measurement, camera calibration, off-axis light source compensation, and light source parameter identification are performed.
Algorithms to calibrate the camera, to scan the object, to smooth out the data, and to construct the 3D model were developed. To test the functionality of the system, and to evaluate the accuracy of the developed system, a controlled experiment is designed and performed. The system works well and a maximum error
of less than 0.01% of full measurement scale is obtained.