Automatic, computer-vision-based identification and assessment of reinforced/prestressed concrete bridge cracks is a problem of interest at the international and national levels. Recently, there have been significant advances in automatic detection and assessment of cracks on bridge decks. Nevertheless, other bridge components are currently subjected to manual-inspection, which exploits the superiority of human vision, tactile, and mobility capabilities over current automatic systems. Hence, cost-effective automation of the crack-detection and assessment process of the entire bridge is extremely challenging. Nevertheless, manual-inspection is expensive and error-prone. In addition, it introduced accessibility challenges and risks to the inspectors. Consequently, numerous attempts to automate the process have been reported. Extensive research, field-testing, and literature review have shown that NONE of the systems reported in the literature, deployed, or tested, throughout the world, can provide an acceptable and cost-effective automatic solution for crack-detection and assessment. Furthermore, accessibility has been identified as a major issue affecting the human performance, accuracy, risk-level, and inspection pace. On the other hand, it is concluded that supervised, semiautomatic systems has the potential to mitigate these issues and provide a highly effective solution to the problem at hand. This paper reports on a Texas Department of Transportation (TxDOT) funded feasibility study that targets the detection, and assessment of concrete bridge cracks. The paper reviews existing systems, experimental systems, and literature; evaluates the challenges of automating the process; and proposes a supervised, semiautomatic approach. The utility of the proposed approach in mitigating the accessibility problem is demonstrated.

K-means clustering is one of the most commonly used methods for classification and data-mining. When the amount of data to be clustered is "huge," and/or when data becomes available in increments, one has to devise incremental K-means procedures. Current research on incremental clustering does not address several of the specific problems of incremental K-means including the seeding problem, sensitivity of the algorithm to the order of the data, and the number of clusters. In this paper we present static and dynamic single-pass incremental K-means procedures that overcome these limitations.