Generation of High Resolution and High Precision Orthorectified Road Imagery From Mobile Mapping System

Abstract

In this paper, a novel technique to generate a high resolution and high precision Orthorectified Road Imagery (ORI) by using spatial information acquired from a Mobile Mapping System (MMS) is introduced. The MMS was equipped with multiple sensors such as GPS, IMU, odometer, 2-6 digital cameras and 2-4 laser scanners. In this study, a Triangulated Irregular Network (TIN) based approach, similar to general aerial photogrammetry, was adopted to build a terrain model in order to generate ORI with high resolution and high geometric precision. Compared to aerial photogrammetry, there are several issues that are needed to be addressed. ORI is generated by merging multiple time sequence images of a short section. Hence, the influence of occlusion due to stationary objects, such as telephone poles, trees, footbridges, or moving objects, such as vehicles, pedestrians are very significant. Moreover, influences of light falloff at the edges of cameras, tone adjustment among images captured from different cameras or a round trip data acquisition of the same path, and time lag between image exposure and laser point acquisition also need to be addressed properly. The proposed method was applied to generate ORI with 1 cm resolution, from the actual MMS data sets. The ORI generated by the proposed technique was more clear, occlusion free and with higher resolution compared to the conventional orthorectified coloured point cloud imagery. Moreover, the visual interpretation of road features from the ORI was much easier. In addition, the experimental results also validated the effectiveness of proposed radiometric corrections. In occluded regions, the ORI was compensated by using other images captured from different angles. The validity of the image masking process, in the occluded regions, was also ascertained.

Object Detection From MMS Imagery Using Deep Learning for Generation of Road Orthophotos

“…A method for generation of road orthophotos using images and point cloud data acquired from MMSs has already been proposed in our previous research (Sakamoto et al, 2012). Firstly, the point cloud of the road surface are extracted from the point cloud data by applying a filtering technique.…”