| The article aims to explore a method to improve the images’ contrast and contourinformation in the visual measurement process. The method adopts parallel lightingsystem, and relies on three cameras capturing images from different angles. Mostimportant is that the resulting image can not only fully inherit a great amount ofinformation from the multi-image, but the contrast of the defects is also greatlyimproved, and the background interference information pixels are effectivelysuppressed, after an efficient and reliable image processing and differential fusion.The above operation is conductive to reducing the complexity of surface defectrecognition and target size measurement.This paper analyzed the reflective properties of the metal surface, and the feasibilityof this topic was proved theoretically. Then a system platform was built, and theexperimental results verified the advantages of differential fusion algorithm inimproving image quality. The main work done in this paper includes:1. Introduced several key technologies in the field of visual measurements, inparticular the lighting technology which played an important role in visionmeasurement system.2. Bidirectional reflectance distribution function and two surface models wereintroduced, and a detailed description of several commonly used surfacereflectance model was done. Then present a surface reflection model applying toour scheme, and predicted the distribution of the reflected light from the metalsurface.3. Taking the standard surface roughness sample block as the experiment object,researched the influence of some factors on the reflected light, such as theincident angle of light, the surface roughness, and surface processing mode. Theone hand, the experiment proved that the differential fusion, based on bright anddark field images, is feasible in theory. On the other hand, the experiment resultsguided the spatial relationship of the various parts in the visual system.4. According to the characteristics of the image acquisition mode in the visualsystem, this paper adopt DLT algorithm combined with RANSAC. Thiscombination calculates fast, keeps an outstanding anti-interference ability, and performs well in image registration accuracy.5. A vision measuring platform was set up, and some surface defects detectionexperiments were implemented on a certain metal work piece. Results show thatdifferential fusion algorithm has a nice performance in improving image contrastand inhibiting interference information. |
