November 03, 2020: Dr. Dinh Tran Hiep (AVITECH), Summit Navigator automatic thresholding for image binarization with application to crack detection

Machine vision has found many industrial applications. For surface inspection, a straightforward and effective segmentation algorithm is of vital importance for a successful extraction of crack pixels from the image background. This work presents a novel binarization method using Summit Navigator for local maxima extraction and contrast-based region merging without any prior knowledge about the test image under segmentation. Based on the detected peaks of the image histogram, an initial segmentation is generated, and a contrast-based region merging technique is proposed to effectively assign different regions into object and background. Then, a bagging technique using the decision trees method is employed to train a classifier for automatic parameter selection. Experimental results on some benchmark data sets demonstrate the advantages of the proposed method over existing techniques in terms of accuracy and consistency. To illustrate its validity, the approach is then applied to detect surface cracks in a commercial building by using images captured by an unmanned aerial vehicle.

Speaker: Dr. Dinh Tran Hiep, AVITECH

Time: 15:30, Tuesday, November 03, 2020

Venue: G2-315, 144 Xuan Thuy, Cau Giay, Hanoi