Application of digital image processing technique in the microstructure analysis and the machinability investigation

OBRABOTKAMETALLOV Том 23 № 4 2021 TECHNOLOGY Microstructure characteristics, i.e. grain size, are measured on a micron (or millimeter) scale. Qualitative and quantitative data become available from the microstructure. This indicates the importance of analyzing microstructural images [3, 4]. In a decision-oriented application, when segmenting an image, pixels can be accurately classi fi ed into several different multiple segments[5, 6]. Image segmentation divides an image into several discrete regions based on pixel similarity. There are many applications of this method, including medical image processing, healthcare, traf fi c image processing, metallurgical industry, pattern recognition, etc. [7–9]. There are many methods of image segmentation, including clustering-based, neural network-based, threshold-based, edge-based, etc. Considering user-friendliness and reliable results, better image segmentation is generally performed by clustering method including k -means, fuzzy c -means, and subtractive clustering, etc. [10] K-means clustering algorithm is one of the best choices for the users. It is simple in execution and faster in computation than other clusterings [11]. It is having the potential to work with a large number of variables and producing different results for different clusters. So, it is essential to start with the proper number of K -clusters. After that, it is essential to start with k – number of centroids. The initial centroid numbers will decide the clusters. So, it is an indication that proper selection of the value of centroid is an essential task [12]. Many methods for the segmentation of color images had been exited. But most of them are application-based. So still there is no universal method for color image segmentation till now. The working of the k -means clustering is shown in fi gure 1 in the form of a fl ow chart. Fig. 1. K-means clustering fl ow chart The purpose of the work is to obtain microstructural quanti fi able information using image analysis and use it to predict the machinability of the material. There is a close relationship between machinability and quanti fi able microstructural information available through image analysis software. There is also a need to apply interdisciplinary approaches in the fi eld of mechanical engineering. Methods The fundamental purpose of image segmentation is to transform an image into an interpreted form for further analysis. But most of the input images are taken from many areas based on the different applications. Some images are not visible, some of them have noise, and some are of poor quality. Therefore they need to be pre-processed before being sent for segmentation [13]. There are so many pre-processing techniques.