Fuzzy C-Means Clustering for Noise Reduction, Enhancement and Reconstruction of 3-D Ultrasonic Images

Resumen

This paper reports the application of artificial intelligence in the reconstruction of images from data acquired via ultrasonic sensors. These elements, placed to form an array of emitters-receivers, take data sequentially from different sections of a piece in movement on a conveyor belt. Although there are innumerable advantages of using ultrasonic sensors in industrial environment, echoes taken from the objects are contaminated with noise, which makes it difficult to properly reconstruct the image. Taking into account this fuzziness (uncertainty) in the measured information and, on the other hand, the search for meaningful regularities that characterize the pattern recognition in image processing, the use of fuzzy clustering algorithms, such as fuzzy c-means, should be of interest. As a comparison, non-fuzzy techniques, such as k-means are also applied, proving to be not as appropriate as the fuzzy alternatives. Another technique related to clustering, the chained distance algorithm, is implemented in order to define the number of regularities or classes in the image to be reconstructed, previous to the clustering. Finally, it is concluded that the use of fuzzy c-means clustering offers excellent results, giving noise-reduced, enhanced images, which are close to the real objects.