Machine learning approach for automated coal characterization using scanned electron microscopic images

• We develop a low cost system for automated coal characterization for preliminary screening of coal samples.
• We utilize the textural and color features in coal sample images to characterize them.
• We create a strategy for achieving better result in terms of accuracy and less computational time.
• To achieve this, we use image analysis technique, pattern recognition, machine learning, etc.

Increased coal utilization has accelerated the need of understanding the basic knowledge of coal quality. Coal is highly heterogeneous in nature and because of its heterogeneity, numerous analytical techniques are needed for its characterization so as to predict its behavior and characteristics. Conventional analysis had been a basic technique long since for coal characterization performed by petrologists. Such conventional characterization of coal samples is time consuming and are limited by the high degree of subjectivity in the results. This paper come up with an automated image analysis approach towards the characterization of given different grades of coal samples. The objective of this work is to improve the characterization of coal samples by analyzing the textural and color features of coal using image processing techniques and to assist in the development of a preliminary screening of the coal samples. Automated characterization of coal is accomplished using image acquisition, features extraction, feature selection and classification over scanned electron microscopic images of coal samples. Hence, authentic and accurate subtyping of coal is obtained with the use of improved prominent features and a standard neural network classifier.