A STUDY ON EXTRACTION METHOD OF SPOTTED SURFACE DEFECTS BY USING DEEP LEARNING FROM CORRECTED 3-D POINT CLOUDS PROFILE DATA

Abstract

 Porous asphalt becomes standard surface mix on the expressways in Japan. This type of surface tends to face spotted defects, such as pumping, alligator cracks lead to potholes, unlike conventional dense asphalt surface, caused by stripping phenomena in the base course mix. This feature of damage causes difficulty of the use of traditional measurement to detect the place, because of its location spot by spot. Therefore, we have been developing a new extraction method and showing its possibility, by using converted road surface measuring data into 3-Dimension clouds profile data. However, proposed method takes rather long time to analyze, not enough capable to evaluate all of surfaces on the existing expressway networks. Hence, further study is conducted to shorten the analyzing time to detect the spotted damage on the porous asphalt surface, by deep learning method with the converting corrected 3-D clouds profile data into color information image of multi–stages.