Keeping core parts of machines in proper condition is essential to improving productivity and quality of products. Metallic wear of knitting needles for circular knitting machines should be controlled within specific conditions. Currently, inspections of them are visually performed by skilled examiners, and automated inspection systems, which can measure 3-D shapes, are demanded. Because the needles have mirror glossed, complexly shaped surfaces, conventional lighting method, such as dome lights and diffuse on-axis lights, cannot irradiate the light evenly throughout the object and causes brightness unevenness, and that leads to 3-D measurement errors in image processors. To increase accuracy, we propose a new 3-D measurement system which equips omnidirectional EL (electroluminescence) lightings and DEHF (Dynamic Enhancement of High Frequency) method. Here, the system applies the shape from focus method, which moves the optical system vertically with respect to the fixed object and obtains a sequence of images that correspond to different levels of object focus. In the formation process of shallow depth of field microscopic images, a defocused imaging system plays the role of a low-pass filter. For this reason, the regions with high frequency components can be regarded as a focused area. The high frequency components are finally regarded as the contour of the object by the method. It recovers the 3-D shape of the object by estimation of height of the contours each image and arranging in the original order of the sequence of the height of the contours. The followings are novelties of our proposed system.
Firstly, omnidirectional EL lightings irradiate an object with uniform lights from all directions. They are composed of the following lights: coaxial through objective lens, object lens perimeter, side and bottom lights, and each of which can adjust brightness; therefore they can reduce unevenness of brightness on the object. We adopted inorganic EL sheet as the lighting device. EL sheet is capable of plane emission and prevent the occurrence of the unevenness of the irradiated light by the point source of light.
Secondly, algorithm for shape from focus can be improved by our DEHF method. Even if the above lightings are applied, there still remains low frequency non-uniformity of brightness. DEFH method removes the low frequency by subtracting mean filtered image from original one, and remaining high frequency content can be emphasized.
We built a microscope based prototype system and conducted experiments. Through them, the validity of our proposed method was confirmed.