نوع مقاله : مقاله کامل پژوهشی
نویسندگان
1 گروه مهندسی عمران - دپارتمان راه و ترابری - دانشکده مهندسی - دانشگاه فردوسی مشهد - ایران
2 گروه عمران، دانشکده مهندسی، دانشگاه فردوسی مشهد
3 گروه مهندسی عمران - دانشگاه فردوسی مشهد
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Assessment of road surface distresses is one of the most prominent phases in the pavement evaluation process. Over the past few years, a considerable number of efforts have been made on developing full-automatic methods for objectively pavement distress detection. It should be noted that most of these methods are based on computer vision techniques. One of the most important assets comprising computer vision systems is the feature extraction procedure. Multi-resolutional analysis approaches, such as wavelets have provided robust tools for image textural features representation. Edges are the textural structures of the pavement distress image. The applied wavelet decomposition level (scale) plays an effective role in highlighting and spatial localizing of the image edges (discontinuities) texture and must be selected optimally. In the present study, after acquisition of asphalt pavement distress images in an illumination controlled condition, three level Haar discrete 2D wavelet transform was applied in order to decompose the images. Afterwards, gray level co-occurrence matrix and gray level run-length matrix were employed to statistically describe the frequency sub-bands texture. The distress classification results based on minimum Mahalanobis distance classifier, indicate that extracting sub-bands statistics based on two layer wavelet decomposition has superior performance in distresses texture recognition. Average classification accuracy rates based on one, two and three levels wavelet decomposition are 68%, 79% and 64%, respectively. Furthermore, textural measurements obtained from gray level co-occurrence matrix, yielding average classification accuracy of 81%, acquired superior results in pavement distress images discrimination compared to gray level run-length matrix statistics.
کلیدواژه‌ها [English]
)