A neural network-based method for coverage measurement of shot-peened panels

Abstract

Shot peening is a cold metal working process to improve the material strength, reduce corrosion fatigue and prevent fracture. Measuring the coverage level is an essential parameter in shot peening, which is traditionally performed through manual visual inspection. Due to the tedious nature of the task, it is prone to imprecision caused by human error. Several image processing and computer vision techniques are proposed in the literature to automate this process. While most of the techniques are accurate in segmenting the shot-peened areas, they seem to fail in the presence of machining streaks, resulting in false segmentation. To overcome this challenge, an artificial neural network (ANN)-based implementation is employed in this paper to improve accuracy of the results. The neural network is trained with specific selected features from the acquired images. Results show ANN outperforms the previously implemented standard image segmentation methods.