基于特征组合的舰船目标粗识别方法研究

Research on Ship Rough Recognition with the Features of Combination

针对单一特征量无法准确识别目标的缺点,文中利用特征组合方式进行目标粗识别。首先,利用基于图相似性分割方法提取目标的二值图像;然后,计算目标仿射不变矩并利用分类器进行聚类分析;最后,通过最小外接矩形法提取目标几何参数,并依据图像分辨率计算出实际目标参数,且与舰船目标配属库进行比对,实现目标粗识别。实验结果表明,该方法鲁棒性强,易于实现,通过实测数据可以识别出舰船目标类别。

Aimed at the shortcoming that single feature could not accurately identify the target,a method for research on ship rough recognition with the features of combination was proposed by this paper.Firstly the binary image of the target was extracted by graph portioning active contours（GPAC） segmentation.Secondly calculated the affine invariant moment of the target and used the classifier for cluster analysis.Finally extracted the target geometry parameters through the Minimum Enclosing Rectangle method,and calculated the actual target parameters according to the image resolution,and comparing with ship target configuration database to achieve the rough recognition of ship target.The experiment results showed that this method had strong robustness and could berealized easily,which could identify the ship target categories by the actually measured data.