Volume XLII-2
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2, 573-577, 2018
https://doi.org/10.5194/isprs-archives-XLII-2-573-2018
© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2, 573-577, 2018
https://doi.org/10.5194/isprs-archives-XLII-2-573-2018
© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

  30 May 2018

30 May 2018

OBJECT DETECTION FROM MMS IMAGERY USING DEEP LEARNING FOR GENERATION OF ROAD ORTHOPHOTOS

Y. Li, M. Sakamoto, T. Shinohara, and T. Satoh Y. Li et al.
  • PASCO CORPORATION, 2-8-10 Higashiyama, Meguro-ku, Tokyo 153-0043, Japan

Keywords: Object detection, Vehicle and its shadow detection, Deep learning, Faster R-CNN, Road orthophoto, Mobile mapping system

Abstract. In recent years, extensive research has been conducted to automatically generate high-accuracy and high-precision road orthophotos using images and laser point cloud data acquired from a mobile mapping system (MMS). However, it is necessary to mask out non-road objects such as vehicles, bicycles, pedestrians and their shadows in MMS images in order to eliminate erroneous textures from the road orthophoto. Hence, we proposed a novel vehicle and its shadow detection model based on Faster R-CNN for automatically and accurately detecting the regions of vehicles and their shadows from MMS images. The experimental results show that the maximum recall of the proposed model was high – 0.963 (intersection-over-union > 0.7) – and the model could identify the regions of vehicles and their shadows accurately and robustly from MMS images, even when they contain varied vehicles, different shadow directions, and partial occlusions. Furthermore, it was confirmed that the quality of road orthophoto generated using vehicle and its shadow masks was significantly improved as compared to those generated using no masks or using vehicle masks only.