An algorithm for indentation image classification and detection based on deep learning

Abstract In the measurement of hardness value, indentation is the key to obtain the hardness value. Traditional manual detection of indentation is inefficient, time-consuming and low accuracy. And some automatic methods can only achieve single indentation detection. In the early stage of full-automatic detection, the hardness block classification and some hardness surface detection with interference impurities need to be carried out manually, which reduces detection efficiency. Therefore, this paper presents an automatic classification and detection method of hardness indentation based on Alexnet neural network. By classifying the hardness indentation first, and then using different analysis methods for different indentation, the detection efficiency can be improved. The experimental results show that the average classification accuracy of Alexnet is 90.75% and the detection speed of a single image is 28.26 ms. This method is advanced and effective in indentation classification and detection.