Building Detection For Polsar Image Based On A Hierarchical Cognition Model

PolSAR systems can work in all weather and all time withmulti-polarization modes, and have wide applications in the field of remotesensing. By observing the scattering information of objects in differenttransmitting and receiving polarization states, PolSAR can fully acquire thecharacteristics of objects. Due to the special imaging method of PolSAR system,PolSAR images have different features from images of other sensors. As aconsequence, PolSAR image interpretation technique, which includesclassification, target detection and recognition and others, is a hot and difficultissue in the world in recent years and is still not yet mature. Target detection isan important aspect in the research of PolSAR image interpretation. Therefore, todevelop PolSAR image target detection technologies is of great value andsignificance to improve PolSAR image interpretation techniques and to promotethe development of PolSAR systems.The purpose of this paper is to develop a new building detection method ofPolSAR images based on human’s cognitive process. Human has a perfect imageprocessing system which can fast and accurately identify interested targets froman image. Studying human’s cognition mechanism and introducing cognitivetheory to the field of PolSAR image target detection can contribute to theaccurate and fast detection of interested targets in PolSAR images. Human’scognition system has two main characteristics. First, human has a hierarchicalinformation processing structure which can process information in an orderedand efficient step-by-step way. Inside the same layer, information is parallelprocessed and in different layers, information is serial processed. Second, thewhole cognitive process is guided by rich experience and knowledge. Therefore,based on human’s cognition mechanisms, considering the characteristics ofPolSAR images, a hierarchical building detection model for PolSAR image isproposed. The proposed model has two levels which are “visual cognition” leveland “logic cognition” level, respectively, and include priori knowledge. Theproposed method provides a new method, a new idea for PolSAR image target detection and has applicable value.This paper first illustrates the research background of related fields,including cognition theory and model, PolSAR image target detectiontechnologies and PolSAR information extraction techniques. Then some basictheories of cognition science and PolSAR techniques are introduced. Based onthese theories, the hierarchical cognition model for PolSAR building detection,which consists of “visual cognition” and “logic cognition” level and includespriori knowledge, is then proposed and introduced in detail. The “visualcognition” and “logic cognition” levels are used to simulate the process of humanvisual perception and logic process, respectively. Fuzzy logic, neural networkand some other theories are applied to accomplish the mathematicalrepresentation and computer implementation of the two levels. In the “visualcognition” level, appropriate primary features are extracted from PolSAR imagesand then the primary features are integrated to derive interest map under theguidance of priori knowledge. Then building detection is done based on theinterest map and scale filter and shape judgment is utilized to remove falsealarms in the “logic cognition” level. The priori knowledge is managed andstored independently in the system, for the convenience of the improvement ofthe system. At last, three different PolSAR images acquired by different sensorsare utilized to validate the proposed method and the results show that theproposed method can effectively detect building areas from PolSAR images andhas some applicability.