Research On Recognition System Of Typical Artificial Target From Ocean Remote Sensing Images

With the development of National Maritime Strategy from offshore to ocean and from regional to global,China urgently needs to enhance global maritime awareness.Real-time fine detection and recognition of ocean-going targets is still a bottleneck in this field.At present,there are still many difficulties in the detection and recognition of typical man-made targets such as offshore oil drilling platforms and ships in China.Relying on the combination of modern remote sensing technology and computer vision technology,the joint detection and recognition technology has become an effective way to solve the above bottleneck problems.Generally,the sea target recognition algorithm of remote sensing image is based on smallscale image,which has high complexity and can only recognize one kind of target.Aiming at the problems of strong pertinence,high complexity and poor portability in the existing recognition algorithms for large-scale visible remote sensing images of typical man-made targets in the ocean.This paper focuses on the recognition of ocean typical man-made targets in large-scale visible remote sensing images,and studies the mask segmentation of remote sensing images,the detection and extraction of regions of interest,the extraction and classification of target features in order to realize the detection and recognition of ocean targets in remote sensing images.The effectiveness of the proposed scheme is verified by experiments.The key technologies studied in this paper are:(1)Aiming at the interference of obstacles such as land and cloud in remote sensing images,an adaptive image segmentation and mask algorithm based on maximum inter-class variance method is proposed,and the results of segmentation are refined by using morphological operations of related images to achieve efficient and accurate mask of obstacles;(2)saliency objects.Standard detection is an important link in the positioning of typical man-made targets on the sea in remote sensing images described in this paper.In this paper,Laplace of Guassian speckle detection algorithm is used to detect the initial response of the target location,but it is difficult to accurately determine the target location.In order to achieve the accurate location of the target,a multi-cluster center detection method based on graph theory is proposed to achieve the accurate location and scale estimation of the multi-target,which is based on the location information and scale information of the target.Information can effectively extract regions of interest;(3)Aiming at the problem of sparse samples and non-specific shape features of some targets in the classification and recognition of remote sensing images,a comprehensive classification scheme based on shape and texture features is proposed to realize the recognition and classification of targets.In order to verify the validity and reliability of the scheme described in this paper,a large number of high-resolution remote sensing images were downloaded as experimental data to test the whole system,and the test results were analyzed in detail.Experiments show that the scheme proposed in this paper can achieve accurate and fast positioning and recognition of man-made targets on large-scale remote sensing images of ships and drilling platforms at sea.The purpose of the study is expected.