| With the development of cloud computing technology and the rapid growth of multimedia data,remote sensing image using cloud environment resources for storage and computing has become a trend.More and more resource-constrained users choose to upload the data to the cloud server for analysis and processing,but the data stored in the cloud server is often exposed to the risk of leakage.In view of this problem,remote sensing image data needs to be encrypted before being outsourced to the cloud.However,encryption operation disturbs the original distribution state of data,which directly affects the normal processing of data and has an impact on the relevant services provided to users,the most typical one is remote sensing image retrieval.Traditional content-based image retrieval(CBIR)is based on the similarity of image data to achieve the image retrieval,but the encryption operation leads to the change of the distance between features,so CBIR cannot be directly applied to the ciphertext remote sensing image retrieval in the cloud environment.In this environment,the research related to remote sensing image safety retrieval is gradually emerging.Due to the rich information and complex background of remote sensing image features,the number of features is huge.At present,the safety retrieval method of remote sensing image generally has the problem that the retrieval accuracy,retrieval efficiency and security of feature data cannot be balanced,which cannot meet the application demand of high efficiency,high precision and high security for mass remote sensing image retrieval.Focusing on the security,accuracy and real-time of remote sensing image retrieval in the cloud environment,this paper makes an in-depth study on image content high-semantic hierarchical expression and encryption in remote sensing image retrieval,and proposes a secure retrieval algorithm for remote sensing image in cloud environment,which can express the content of remote sensing image at a higher semantic level while taking into account the security of retrieval,so as to further improve the retrieval accuracy and efficiency.The main contributions of this paper are as follows:(1)Remote sensing image data has a high degree of redundancy and a strong correlation between the data.The traditional encryption method is not safe when applied directly to image feature encryption.Aiming at the problem of abundant redundant information in remote sensing image,this paper studies the feature screening method of remote sensing image.Considering the relationship between the characteristics of similar and different images,this paper proposes a feature selection method based on CHI square test(CHI)to select the visual features of remote sensing image adaptively,which can reduce the redundancy of remote sensing image data and provide support for efficient and high precision retrieval of remote sensing image.The effectiveness of the method is verified by experiments.(2)Aiming at the complex background and abundant information of remote sensing image,this paper studies the security expression method of high-level semantics of remote sensing image and proposes a secure retrieval method of remote sensing image based on dictionary optimization to realize the efficient and high precision retrieval of remote sensing image.BOVW model and Markov reinforcement learning model are combined to learn the relevance between visual codebooks,which can achieve the high-level semantic expression of remote sensing image and improve the retrieval accuracy.Hamming embedding and min-hash algorithm are used to construct the security index of remote sensing image and to ensure the efficient and high-precision security retrieval of remote sensing image.The effectiveness and feasibility of the method are proved by theoretical analysis and experimental verification.The method proposed in this study realizes the efficient and high-precision and highsecurity retrieval of remote sensing images in the cloud environment,and ensures the efficient and safe use of remote sensing images in the open cloud computing platform. |
