In recent years, autonomous mobile robots have increasingly been introduced not only into manufacturing plants but also into human living environments. One important means for such robots to understand their surrounding environment and their own posture is the range image sensor, whose importance continues to grow. Such sensors are expected to provide highly accurate measurement as well as high-speed real-time performance. Based on previous work that reduced the correspondence problem by using a multi-slit laser, this study aims to realize high-accuracy, high-speed real-time measurement.

Three-dimensional measurement based on image processing is actively studied in a wide range of fields, including medicine, automobiles, factories, and construction. Stereo cameras are often used in these studies. However, stereo cameras suffer from reduced accuracy in textureless environments. In this research, we propose a method that uses a line laser as an auxiliary device for a compact stereo camera, enabling high-precision three-dimensional measurement even for textureless objects.

In this research, range images are acquired using a monochrome camera and a laser projector that projects 15 slit-light patterns. In previous studies, the measurement range had to be limited in order to avoid the correspondence problem. We are investigating a method to mitigate this problem and expand the measurement range by identifying individual slits using defocus.

By overlaying multiple range images acquired from a range image sensor, it becomes possible to generate a three-dimensional map of a scene. In this research, the motion parameters of the range image sensor are estimated using constraint equations based on range images and color images. The estimated motion parameters are then used to align the range images and generate a three-dimensional map. Future work will focus on improving the accuracy and speed of three-dimensional map generation.