For Project Enquiry +91 9962 067 067

EXISTING SYSTEM

PROPOSED SYSTEM

EXISTING CONCEPT:

         This paper proposes a novel hierarchical complete and operational SDSOI approach based on shape and texture features, which is considered a sequential coarse-to-fine elimination process of false alarms. First, simple shape analysis is adopted to eliminate evident false candidates generated by image segmentation with global and local information and to extract ship candidates with missing alarms as low as possible.

PROPOSED CONCEPT:

           According to the recent works in, deep architecture has multiple levels of feature representation, and the higher levels represent more abstract information. From a concept point of view, DNN trains multiple hidden layers with unsupervised initialization, and after such initialization, the entire network will be fine-tuned by a supervised back propagation Algorithm.

EXISTING  TECHNIQUE :

         LOCAL MULTIPLE PATTERNS (LMP),LOCALBINARY PATTERNS (LBP)

PROPOSED ALGORITHM:

         COARSE SHIP LOCATING, SHIP FEATURE REPRESENTATION AND CLASSIFICATION

TECHNIQUE DEFINITION:

          LMP is introduced to enhance the representation ability of the feature set.

          The LBP operator is a robust but theoretically and computationally simple approach. It brings together the separate statistical and structural approaches to texture analysis, for the simultaneous analysis of both stochastic micro textures and deterministic macro textures.

ALGORITHM DEFINITION:

          Fast ship locating (i.e., ship candidate extraction) is performed in LL subband, which includes image enhancement, sea–land segmentation, and ship locating based on shape criteria.

          The state-of-the-art ship detection approaches extract complicated features and combine them with learning-based classification.

DRAWBACKS:

          First, missing detection exists when a part of a ship is covered by a large cloud, when a ship adjoins a large island, or when the gray of a ship is very close to that of its neighbor. Local segmentation and matching may be a solution.

          Second, false candidates, which mainly comprise clouds and sea clutter, also exist. More effective features are needed to distinguish between them.

ADVANTAGES:

          Multiresolution analysis and singularity detection

         Faster detection. Compressed domain achieves much faster detection than pixel domain.

           More reliable results. High-level feature representations are extracted by hierarchical deep architecture to ensure more accurate classification.

           Better utilization of information. Two DNNs are trained with multisubbands coefficients to make full use of the wavelet information

          Our approach performs better in extracting, training, and testing time, and these results benefit from all aforementioned advantages: faster feature extraction and higher learning efficiency of ELM.

          The main advantage is that the method needs no tuning and can even cope with scenes that are difficult to analyze for a human operator.