Objective
Detection and analysis of low contrast flaws in radiographic images with high noise fields is a topic of current research in non-destructive evaluation of materials and articles during diagnostics. This problem is almost the same in medical diagnostics from radiographs for detection of dangerous diseases. Two crucial applications are considered: technical diagnostics of welds in nuclear power engineering and medical diagnostics of lung cancer. The poor quality of radiographic images is due to the physical nature of radiography as well as small size of the flaws and their poor orientation relatively to the size and thickness of an evaluated part. The known methods for radiograph analysis fail to detect low-contrast flaws and to make their meaningful analysis at all. The main approach to the radiographs processing consists of a development of structure-adaptive methods which consider shape constraints of local objects (flaws) in order to preserve local objects during filtering or to detect them on a non-homogeneous background. On one hand, these methods should be immune to the presence of noise and robust in conditions of non-homogeneity of the background. On the other hand, they must be computationally simple for possible implementation in real time. For flaws detection and analysis, a model based approach is used which provides image analysis by hypothesis generation and testing based on a current state of analysis and a database of hypotheses which was constructed during the learning stage. This is a process of sequential detection of local objects which are formed as structures of several primitive patterns which are described by their planar shape and have different locations in space at different resolution. For correct interpretation of the radiographs, formal decision rules will be used which consider, besides shape features, other image local properties. To evaluate basic image properties involved in the statistical hypothesis testing, an adaptive method of filtering will be investigated based on underlying structural mathematical model of radiographs. The proposed model of radiographs consists of separate morphological model of local objects (flaws) and non-stationary polynomial model for the background. Since the algorithms for image analysis are rather complex procedures, important concern of this research is fast implementation possibility of image analysis which can enable to do real-time diagnostics. For time-efficient processing of radiographs, fast rrsive algorithms with spatial and temporal rrsions will be developed yielding a speedup of an order as compared to the direct implementation of some algorithms. During the research period, the developed algorithms will be incorporated as a software package into available modern equipment for technical and medical diagnostics in the mentioned applications.
Topic(s)
Data not availableCall for proposal
Data not availableFunding Scheme
Data not availableCoordinator
5020 Salzburg
Austria