State and local maintenance departments are tasked with keeping roads in good repair, which includes monitoring the roads to detect the presence of cracks, potholes, and other distress. Currently this is done manually by inspectors, with specialized vehicles or through citizen reports. These methods are tedious, expensive, or unreliable. We have developed a system to monitor road distress on a continuous basis, at low cost, with consistent quality, and with minimal human intervention. Our approach involves using images or videos collected by commodity devices such as smartphones to obtain information about the road condition. These devices can be mounted in service-vehicles like cabs, garbage trucks, police cars, etc., which already make regular rounds on city roads for other purposes. Therefore no dedicated vehicles or drivers are needed, which significantly reduces the cost of the system. Machine vision algorithms analyze the images automatically and can pass the resulting distress scores to the asset management system.Another capability of our system is to find stop signs and determine if they have problems like defacement, occlusion, or damage. We have already started to extend the system to determine road conditions (snow, slush, etc.) and to inventory and assess other infrastructure like lane markers.
