Ongoing Studies for Automatic Road Anomalies Detection on 2D and 3D Pavement Images

Autor: KADDAH, W, El Bouz, Marwa, Ouerhani, Yousri, Alfalou, Ayman, Desthieux, Marc
Přispěvatelé: Light Scatter Learning (LABISEN-LSL), Laboratoire ISEN (L@BISEN), Institut supérieur de l'électronique et du numérique (ISEN)-YNCREA OUEST (YO)-Institut supérieur de l'électronique et du numérique (ISEN)-YNCREA OUEST (YO), Kaddah, Wissam
Jazyk: angličtina
Rok vydání: 2018
Předmět:
Zdroj: The International Symposium on Optoelectronic Technology and Application (OTA)
The International Symposium on Optoelectronic Technology and Application (OTA), 2018, Beijing, China
Popis: International audience; A pavement is a solid surface material laid down on an area intended to support foot and vehicular traffic, such as a walkway or road. Due to heavy traffic and poor weather conditions, the road becomes increasingly degraded and thus threatens human safety. For this reason, automatic road anomalies detection is introduced in order to solve problems of the pavement and avoid the expensive maintenance operations. In this context, we will present in this paper two types of anomalies we have considered to achieve the main goal of our research. The first type concerns the detection of road marking degradations. To do so, we are based on our own method cited in [1] for road marking features extraction using the VIAPIX® system [2]. Indeed, relying on inverse perspective mapping technique and color segmentation to detect all white objects, our algorithm is able to examine images automatically and to provide information on road marks. Then, based on an optical correlation and a geometric recognition techniques to identify the detected objects and a technique for analyzing the state of the identified marks, it allows qualifying all road markings more accurately with a minimum of false alarms. The second type of anomalies concerns fine-structures extraction defined by our novel approach ADFD (Automatic Darkest Filament Detection) for automatic road crack detection on pavement images acquired by the Aigle-RN system [3, 4]. Indeed, our approach is composed of two main phases. Relying on the photometric characteristics, the first phase consists in selecting dark pixels that have great probability of belonging to a crack in the image. Then, based on the geometric characteristics, the second phase consists in connecting the selected dark pixels between them by applying the Dijkstra's algorithm [5] in order to get the crack skeleton in the pavement image.
Databáze: OpenAIRE
Externí odkaz: