Automatic inshore ship detection from remote sensing imagery has many important applications

such as ship change detection and harbor dynamic surveillance. Stable performance of inshore ship detection is vital to the analysis of ship change and the determination of the harbor surveillance effect. Ship detection using optical remote sensing images has been a hot research topic. However

detecting inshore ships utilizing the traditional area-based method is difficult because the gray scale and texture character of inshore ships are similar to that of the shore. Therefore

we propose a method of inshore ship detection using local salient characteristics. First

the binary image is obtained by water and land segmentation preprocessing. Then

the line segments from binary images are extracted as the local salient features to detect ship targets. Next

line segment extraction result is combined with ship bow detection result to generate the ship detection model. Finally

the ship targets are acquired by calculating the ship geometric size and analyzing the environmental information. Experimental results indicate that the proposed inshore ship detection method is more effective and can robustly adapt to the complex background and mooring orientation. The detection result is more accurate compared with that of traditional methods

and the relognition rate is 100%. In the complex background environment and other interferences

this method exhibits a high recognition rate

with high robustness and high adaptability.