Visual method of road damage assessment at railway and highway crossing

Assessment of road condition can be done by visually assessing the level of road damage and also measuring road deflection. Currently measuring road damage assessment uses a visual method with direct observations in the field. The results of the assessment of the level of road damage are calculated from the value of each type of damage. Visual assessment of the level of road damage can be used as a basis for determining the handling of road repairs in the form of road maintenance and road improvement. There are several methods of visually assessing road damage, such as the Bina Marga method and the Pavement Condition Index (PCI) method. Several methods of assessing road damage using the visual manual method so far require the accuracy of surveyors to determine the type and level of damage. The level of accuracy of each surveyor may vary depending on experience in assessment. In addition, the use of this method takes a long time to assess road damage. Road damage assessment needs to use other better methods. One of them is by using data from image sensors, so that road damage data can be obtained more accurately and quickly. Therfore, it is necessary to propose the use image sensors to help asses road damage specially at Railway and Highway Crossing. This papper will review several methods of visually manual road damage assesment and plans for using image sensor on vehicle with the Convolutional Neural Network (CNN) method based on the Edge Tensor Processing Unit (TPU) with the MobileNet SSD v2. Preliminary research results indicate that the use of image sensors will show high accuracy values at vehicle speeds up to 20 km/hour and measurements are made during the day.