NDE data fusion using morphological approaches

Abstract

The objective of most data fusion algorithms is to combine information made available by various sensors synergistically in order to enhance the overall level of information. Since information obtained from data sources such as sensors is often incomplete or imprecise in nature, the application of data fusion techniques has evoked interest in a number of fields ranging from robotics to nondestructive evaluation (NDE). In NDE applications, such techniques can be used to integrate and fuse data obtained using multiple inspection modalities to produce a more comprehensive picture of the condition of the test specimen. As an example, ultrasonic and eddy current imaging techniques are used very widely to inspect a variety of materials. Each technique offers inspection capabilities and limitations that are dictated by the underlying material/energy interaction process. The information generated using the two methods can be construed either as complementary or redundant in nature. Ideally it should be possible to utilize the redundant information to improve the signal-to-noise ratio. Likewise, it should be possible to fuse the complementary information from the two tests to increase the overall level of information made available to the analyst. Unfortunately the task of segmenting data as noise, redundant and complementary components of information can be frustrating. Consequently, most of the approaches proposed to date in NDE have relied on alternate methods;This dissertation proposes a new algorithm for fusing ultrasonic and eddy current images employing morphological imaging processing approaches. The fusion is accomplished in two stages. The first stage basically employs morphological approaches to reduce unwanted artifacts such as speckle noise in the ultrasonic image. The second stage extracts information about the locations and boundaries of defects on the basis of information contained in the morphological granulometric size distribution of the ultrasonic image. Data fusion is accomplished by combining information relating to the locations and boundaries of the defect obtained from the ultrasonic data with the defect depth information derived from the eddy current image. The validity of the approach is demonstrated using several experimentally derived ultrasonic and eddy current images.