This thesis presents work carried out on the development of a real-time variable rate bitumen spraying system. Sealing roads with bitumen binder and stone chip is used to provide a protective, wear resistant layer and waterproof road surfaces. One of the predominant causes of road surface failure is flushing, which is the bitumen binder bleeding through the aggregate, and is caused by an excess of bitumen binder. The build-up of excess bitumen binder replicates itself on successive re-seals, with flushed areas bleeding through to the next layer, exacerbating the problem. Different parts of the road surface require different bitumen application rates based on the applied loading in service, and the amount of flushing present. Using a single bitumen spray rate, the seal design is a compromise between an application rate that suits the high loading in the wheel paths, with some possible chip loss elsewhere (between and outside wheel paths), and an application rate that suits the majority of the road.

Transverse Variable Spray Rates (TVSR) are varied across a full lane width to apply more bitumen binder in between the wheel paths and on the shoulders, and less in the wheel paths. The current methods of variable rate spraying typically use a constant profile along the length of the road and do not adapt to specific road conditions. Other options involve using a road network survey vehicle that collects and stores the surface texture using data gathered from a network survey vehicle. The data is then fed to a separate vehicle and the spray rate is adjusted according to the data and GPS coordinates. This multi-step process leads to a great deal of inaccuracy through the use of multiple vehicles and the poor resolution afforded by GPS. The ability to measure the road surface texture and vary the spray rate in real-time using a single vehicle will greatly improve the accuracy and efficiency of application. The benefits of this are two-fold: a reduction in cost through minimising bitumen waste, and producing longer lasting reseals.

A new, simplified, and low-cost method is developed to obtain surface texture measurements in real-time that enable further development of a sensing and variable-rate bitumen spraying system in a single vehicle.

The aim of this PhD was to develop image processing algorithms and an initial proof-of-concept prototype for a real-time variable rate bitumen spray system. A number of methods for measuring the current road surface texture were investigated. Due to the high cost of commercially available systems, and the lack of need for the high rate of data collection offered, a custom machine vision system was developed specifically for this project. The machine vision system utilises a custom-made laser line generator and a commercially available industrial USB 3.0 camera to capture and transfer image data to a processing computer. The image data is then processed using novel algorithms developed specifically for this project to extract the road surface texture information from the images. The output result is a discretised measure of surface texture, which is used to calculate an optimum bitumen application rate for the measured area.

A number of trials were conducted in order to test the efficacy and repeatability of the measurement system. The results from one trial demonstrated the limits of down-sampling the spatial resolution in both axes while retaining key features of the road surface. This result may later be used for determining a suitable spatial interval that matches the bitumen spray nozzle resolution. Dynamic continuous motion in the direction of travel was introduced, along with controlled image capture at the intended spatial intervals in the direction of travel. A trailer test rig was built for the purpose of this study, and was towed by a vehicle along roads identified to have a range of surface textures. Large amounts of testing data was collected and post-processed allowing comparison with visual inspection. Following on from this, a trial was designed to verify repeatability of measurements of the machine vision system. The results showed that the machine vision system produced accurate and repeatable measurements. Finally, a study was conducted to optimise the filter used in the final surface texture measurement algorithm developed during the course of this study. The output results for texture measurement were then compared against the current gold-standard measurement method – the sand circle test. The output metric monotonically increased with the sand circle test, and conformed to expected behaviours.

Additional considerations and future work on the development of the final commercial application of a real-time variable rate bitumen sprayer are discussed, including the introduction of real bitumen spray nozzles to the system. In this study, LED lights were used as a surrogate to bitumen spray nozzles. This addition allowed for the investigation of using a micro-controller to output a physical command to a surrogate component, allowing for the demonstration of precise synchronisation between camera trigger and bitumen binder application events, regardless of changes in vehicle speed, without the need for spraying hot, volatile bitumen solution. Finally, upscaling the prototype to a full-size apparatus to cover both surface texture measurement and spraying of full road lane-widths is discussed.

The developments of this thesis provide the technical groundwork to proceed with the development of economically feasible real-time TVSR for road repair within New Zealand and Internationally.