The task of railroad inspection is a tedious one. It requires a lot of skilled experts and long hours of frequent on-field inspection. Automated ground equipment systems that have been developed to address this problem have the drawback of blocking the rail service during inspection process. As an alternative, using aerial imagery from a UAV, Computer Vision and Machine Learning based techniques were developed in this thesis to analyze two kinds of defects on the rail tracks. The defects targeted were missing spikes on tie plates and cracks on ties. In order to perform this inspection, the rail region was identified in the image and then the tie plate and tie regions on the track were detected. These steps were performed using morphological operations, filtering and intensity analysis. Once the tie plate was localized, the regions of interest on the plate were used to train a machine learning model to recognize missing spikes. Classification using SVM resulted in an accuracy of around 96% and varied greatly based on the tie plate illumination and ROI alignment for Lampasas and Chickasha subdivision datasets. Also, many other different classifiers were used for training and testing and an ensemble method with majority vote scheme was also explored for classification. The second category of learning model used was a multi-layered neural network. The major drawback of this method was, it required a lot of images for training. However, it performed better than feature based classifiers with availability of larger training dataset. As a second kind of defect, tie conditions were analyzed. From the localized tie region, the tie cracks were detected using thresholding and morphological operations. A machine learning classifier was developed to predict the condition of a tie based on training examples of images with extracted features. The multi-class classification accuracy obtained was around 83% and there were no misclassifications seen between two extreme classes of tie condition on the test data. / Master of Science

This research aims to investigate the application of infrared thermography (IRT) as a method of nondestructive evaluation (NDE) for the detection of defects in the rail base area. Rails have to withstand harsh conditions during their application. Therefore, defects can develop in the base area of rails due to stresses such as bending, shear, contact, and thermal stresses, fatigue, and corrosion. Such defects can cause catastrophic failures in the rails, ultimately leading to train derailments. Rail base defects due to fatigue and corrosion are difficult to detect and currently there are no reliable or practical non-destructive evaluation (NDE) methods for finding these types of defects in the revenue service. Transportation Technology Center, Inc. (TTCI) had previously conducted a research on the capability of flash IRT to detect defects in rail base area based on simulation approach. The research covered in this thesis is the continuation of the same project.In this research, three rail samples were prepared with each containing a notched-edge, side-drilled holes (SDHs), and bottom-drilled holes (BDHs). Two steel sample blocks containing BDHs and SDHs of different sizes and depths were also prepared. Preliminary IRT trials were conducted on the steel samples to obtain an optimal IRT setup configuration. The initial inspections for one of the steel samples were outsourced to Thermal Wave Imaging (TWI) where they employed Thermographic Signal Reconstruction (TSR) technique to enhance the resulting images. Additional inspections of the steel samples were performed in the Southern Illinois University-Carbondale (SIUC) facility. In case of the rail samples, the SDHs and the notched-edge reflectors could not be detected in any of the experimental trials performed in this research. In addition, two more rail samples containing BDHs were prepared to investigate the detection capabilities for three different surface conditions: painted, unpainted, and rusted. The painted surface provided a best-case scenario for inspections while the other conditions offered further insight on correlating the application to industry-like cases.A 1300 W halogen lamp was employed as the heat source for providing continuous thermal excitation for various durations. Post-processing and analysis of the resulting thermal images was performed within the acquisition software using built-in analysis tools such as temperature probes, Region of Interest (ROI) based intensity profiles, and smoothing filters. The minimum defect diameter to depth (aspect) ratio detected in preliminary trials for the steel sample blocks were 1.0 at a diameter of 4.7625 mm (0.1875 in) and 1.5 at a diameter of 3.175 mm (0.125 in). For the inspection of painted rail sample, the longest exposure times (10 sec) provided the best detection capabilities in all sets of trials. The three holes having aspect ratio greater or equal to 1.0 were indicated in the thermal response of the painted and rusted samples while only the two holes having aspect ratio greater or equal to 1.5 were indicated in the unaltered sample. Indications of reflectors were identified through qualitative graphical analysis of pixel intensity distributions obtained along a bending line profile. The results obtained from the painted sample provided a baseline for analyzing the results from the unpainted and rusted rail samples. This provided an insight on the limitations and requirements for future development. The primary takeaway is the need for an optimized heat source. Poor contrast in the resulting image for the unpainted and rusted rail samples is experienced due to both noise and lack of penetration of the heat energy. This could have been due to decreased emissivity values. Moreover, the excitation method employed in this research does not comply with current industry standards for track clearances. Therefore, exploration of alternative excitation methods is recommended.