A specular surface reflects light only in the direction such that the angle of incident equals the angle of reflection. Due to this characteristic, conventional approaches to obtain the shape of specular objects may not guarantee accurate results. Taking into account this characteristic of the specular reflection, an optical sensing system is proposed to accurately obtain the 3-dimensional surface shape of specular objects. The sensing system is so designed that highly focused laser beam strikes the surface of object and the specular components of the reflected light are then detected by a beam receiving unit. To achieve this measuring principle the system is composed of a galvanometer which steers the laser beam's direction to the desired surface point, a mirror unit(a parabolic mirror and a conic mirror)which reflects the specular component of surface reflection toward its center and a beam receiving unit positioned along the mirror center line. A series of experiments was performed to evaluate the performance of the proposed sensing system. The sensing principle and measurement accuracy are discussed in detail for various objects of simple geometrical shapes.
A surface of solder joints is a typical example of specular surface. Due to the specularity and the complex three dimensional geometry, the visual inspection of solder joints has been regarded as one of the most difficult tasks and thus has not guaranteed accurate inspection results. If laser scanning unit scans the area of solder joint and observes the angle of the reflected beam, the proposed system is able to obtain the external profile of solder joints accurately. To classify the defects of solder joint, we used two pattern classification methods, a statistical pattern recognition method and a neural network classifier. To verify the validity of the developed system for solder joint inspection, a series of experiments was performed for SOPs and QFPs in insufficient, normal and excess soldering condition. From observation of the experimental results, the proposed system is found to show good performance for inspection of solder joint defects and the correct classification ratios of the neural network classifier are slightly larger than those of the statistical pattern recognition method.