International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 06 Issue: 11 | Nov 2019 www.irjet.net p-ISSN: 2395-0072 © 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 355 SAR IMAGE RETRIEVAL BASED ON REGION BASED SIMILARITY MEASURE FOR EARTH OBSERVATIONS RACHIT JAIN 1 , SUMIT YADAV 2 , IRFAN MINSURWALE 3 , SANDHYA SHINDE 4 1,2,3 Student, Dept. of Electronics And Tele-Communication Engineering, Dr. D.Y Patil Institute Of Engineering, Management and Research, Maharashtra, India. 4 Assistant Professor, Dept. of Electronics and Tele-Communication Engineering, Dr. D.Y Patil Institute of Engineering, Management and Research, Maharashtra, India. ---------------------------------------------------------------------***---------------------------------------------------------------------- Abstract - Based on the region-based similarity measure, a novel synthetic aperture radar (SAR) image retrieval method is proposed in this paper, which is inspired by the existing content-based image retrieval (CBIR) techniques and is oriented toward the Earth observation (EO). First, due to the large sizes of SAR images, new method semantically classifies the land covers in the patch level rather than the pixel level by the classic semi-supervised learning (SSL), which could reduce the workload of selecting the representative image patch and decrease the searching space in the similarity calculation component. Furthermore, to overcome the inevitable classification error, our method provides an error recovery scheme, preventing the errors produced in categorization to contaminate the retrieval results. Third, the similarity between two patches is calculated by the improved integrated region matching (IIRM) measure based on the region-based similarity measure, which fails to meet the expectation in SAR images. The proposed method can be embedded into any EO mining systems to help them complete the EO missions. After comparing the method presented in this paper to others, it is evident that our method performs more effectively than others from the CBIR aspect. Key Words: Content-based image retrieval (CBIR), Earth observation (EO), improved integrated region matching (IIRM) measure, synthetic aperture radar (SAR) image retrieval, Semi-Supervised Learning (SSL). 1. INTRODUCTION Among those systems, Google’s Image Search engine may be the most popular one at present. The truly successful general-purposed retrieval system has not emerged since the issues these systems address are too board, and the well- known semantic gap, i.e., the gap between low-level visual features and high-level semantics, exists in the CBIR technique. However, one of the extensions and utilizations of CBIR methods, RS image retrieval (RSIR), has been made greater success. Numbers of famous and practical RSIR methods have been proposed in recent years. Some of them focused on feature extraction and object semantic representation, and others were concerned on applying higher semantics. There are numerous RSIR methods, while the methods aimed to general-purposed synthetic aperture radar (SAR) images retrieval are rare. As one of the EO products, SAR images have drawn increasing attentions recently due to the special characteristics such as possible of penetrating clouds, and operating in bad weather and night. With the increasingly SAR data open and free, a growing number of application on SAR data would be developed in future years. Thus, automatic and semiautomatic interpretation of these SAR data becomes both important and urgent, such as finding the interested regions of researchers from a large volume of images, understanding the distribution of land covers in one SAR image, and detecting changes of the same region in different times, etc. In this paper, a novel SAR images retrieval method oriented toward the EO mining is introduced, which is based on region-based similarity measure. Different from many existing EO mining techniques, our retrieval method focuses on designing a robust similarity measure rather than using more complex information, such as metadata, to complete the mining mission. To construct the database, SAR images in big sizes are divided into equal sized image patches first, and then users pick up some representative patches. The semi- supervised learning (SSL) would be implemented to label all the image patches in database. Here, we adopt SSL to be classifier because it needs less train samples in categorization, which can ease the burden of the image patch selection. The classification step in our retrieval method could reduce the workload of the representative patch selection, narrow down the searching space in the resemblance calculation component, and decrease the impact of semantic gap to our method. When users input a query image patch, the similarities between the query and relevant target patches (the image patches exist in database) would be calculated by improved integrated region matching (IIRM) measure introduced in this paper for SAR images, and the retrieval results are displayed in order. The time consumption is concentrated on the similarity computation in our retrieval method. Due to the high efficiency of new similarity measure, and the pre -classification, our retrieval method is fast. This paper focuses on the task of SAR image retrieval oriented toward the EO mining. Several improvements are added into traditional CBIR techniques to enhance the retrieval precision. The significant contribution of this work can be summarized as follows.