Ship Detection In Harbor Image Based On Priori Geographic Information

Ship automatic detection technology has a broad application prospect in military and civilian fields,plays a crucial role in specific ports monitor,auxiliary shipwreck rescue and illegal fishing cracking down.Due to port and ships have similar gray texture features,the inshore ship detection has been a difficulty.In this paper,we research about inshore ship detection in optical images shot by aircraft,main work includes the following aspects:First of all,analysis the complicated background of sea and land,and combine with the specific test requirements and test conditions,put forward an inshore ship detection scheme based on the priori geographic information.This solution takes the harbor template matching method to locate the sea area with the guide of priori geographic information,simplify the inshore ship detection to offshore ship detection,and then segment the sea area to detect ships.Harbor matching algorithm research.According to the geometric shapes and typical linear features in harbor,we select the edge feature to match.First,introduce several common edge matching similarity measure,we proposed an edge gradient vector based matching method to overcome the interference from edge in land and improve the matching accuracy and precision.Considering the matching speed and attitude angle error,matching strategy is optimized.We put forward a coarse to fine match process.On the one hand,improve the matching speed in coarse matching;On the other hand,add the attitude angle error correct process to reduce the influence of angle error.And through the experimental data show that the algorithm is effective.Sea-land segmentation research.We proposed a segmentation algorithm which combines feature fusion clustering algorithm and local threshold determine method,to solve hull fracture and land holes phenomenon caused by uneven gray.This method firstly adopts K-mean clustering algorithm to divide image tiles into two categories according to the gray level and variance characteristics;and then take local threshold determine method to further optimize edge segmentation effect.Finally,analysis the ship target feature and design a ship detection scheme.We label the suspected candidate area,separate the abreast ships,fix connected domain and exclude false alarm by ship geometry parameters.