Application of Text-Mining and Image Processing Techniques on Digitizing Drillers Logs and Developing Big Well Log Data Sets

Geotechnical and geological information derived from drillers logs constitutes an important source of data for construction and verification of subsurface models. During the past decades, bulky well registration and drillers log data recorded in paper format have been made accessible as image files. More than 90 percent of these lithologic descriptions in the well registration forms are in handwritten format. Converting these data into a machine-readable format is indispensable for further usage in numerical models. However, manual entering of the recorded data in large quantities is time-consuming and prone to human error. The goal of this work is to develop an automated method to convert all scanned information obtained from the wells to the structured file format that is easily employed for further modeling and engineering analysis. The automated procedure was successfully tested to convert the scanned data into images, crop the images to grab the desired pieces of information using Python libraries, and to convert the data from an image format to ASCII files using pre-trained machine learning libraries in Google Cloud Vision API. The proposed method was proved to be highly promising for efficient management of the large data and significantly reducing data manipulation time.