Acceleration of 3D Measurement of Large Structures with Ring Laser and Camera via FFT-based Template Matching

摘要

For quality inspection and maintenance in the manufacturing industry, it is extremely important to accurately measure the 3D shape of various structures in an efficient manner. In this paper, we propose an accelerated light-section method to accurately measure the 3D shape of large structures. The light-section method is a popular method to measure the 3D shapes accurately, which uses a laser and a camera that can observe the shape of the laser cross-sections. This method consists of two important steps: first, images that include the laser cross-section are captured. Next, the cross-sections are integrated by calculating the transformations between these images. For want of high accuracy, area-based matching methods, such as template matching, which match pixels in an area one-by-one, are used to calculate these transformations. Hence, the processing time is too high and speed/efficiency is quite important for real-life 3D measurement applications. Therefore, we focus on developing a faster template matching method. In this process, the presence of the laser cross-section on the image texture can induce errors in the integration and lower accuracy. Thus, it is required to mask the laser regions inside each image. To solve this problem and make the calculation faster, a masked Fast Fourier Transform (FFT) based template matching is proposed. Via experimental evaluation, we show that the 3D measurement process can be made almost 1.6 times faster, with similar accuracy as compared to previous methods.