基于Radarsat-2数据的渔船检测与识别方法研究

Detection and Identification of Fishing Vessels Based on Radarsat-2 Data

采用Radarsat-2高分辨率雷达数据和实测数据相结合的方式,探索了海上渔船目标的检测和识别方法。本研究采用的方法是,首先针对渔船在图像上的表现特征,采用全局均值与一倍标准差之和替代传统双阈值法中的背景峰值,采用目标的强峰与弱峰的中间值替代目标峰值点,作为分割阈值进行目标分割;同时,根据研究海域船只实际尺寸数据和高分辨率图像(GeoEye)目视解译数据,统计取得渔船的长度范围为8.85m^38.58m,长宽比为2.86~5.80,并以此范围进行渔船识别。最后采取交互式精度评价方法,对检测和识别结果进行评价。

An automatic method for detection and identification of fishing vessels on sea was studied based on Radarsat-2high resolution data and the measured vessel geometric data.According to the geometric features of fishing vessels in the SAR image,the traditional double threshold segmentation algorithm was improved:using the global mean and sum of standard deviation as the pure background pixel segmentation threshold value instead of the background peak and the average value of the target highest peak and lowest peak instead of the single peak value as the target threshold.According to the in situ survey vessel data and the geometric data interpreted from the very high resolution image(GeoEye)in yellow sea,the length of fishing vessels ranged from 8.85mto 38.58mand the ratio of lengthvswidth of fishing vessels were from 2.86to 5.80.Interactive evaluations of the effect of fishing vessel detection and identification and the uncertainty analysis for the method were also applied in this study in order to improve the results of fishing vessel detection and identification.