基于自适应上下文信息的目标相对定位方法

摘要

According to the difficulties of the recognition for the ground stationary target in imaging termi-nal guidance,a new localization and recognition approach based on the maximally stable external region (MSER)is proposed.MSER features which are scale and affine invariant are extracted respectively around the target in the reference image.A certain amount of MSER features are adaptively selected as context landmarks according to the weighting exponent.The matching feature pairs between the MSER features extracted from the real-time image and the context landmarks are obtained based on the normalized cross correlation criterion.The double matching correction strategy is used to reduce the probability of mismatching.The center points taken as the reference points are extracted from the matching feature pairs to solve the space mapping relation between the reference image and the real-time image.The position coordinate of the target in the real-time image is calculated according to a polynomial based on the least squares fitting.In view of the complex ground scene,the experi-ments demonstrate that the maximum relative error location probability of the method is less than 3 pixels.It can satisfy the imaging terminal guidance for automatic target recognition algorithm requirements of higher pre-cision and rapid speed,as well as strong anti-jamming and stabilization.%针对成像末制导中地面固定目标识别难度大的问题，基于最稳定极值（maximally stable external re-gion，MSER）区域提出一种新的相对定位识别算法。提取基准图中目标周围具有尺度和仿射不变特性的 MSER特征，根据权重指数自适应选取一定数量的 MSER 特征作为上下文地标。提取实时图中的 MSER 特征，与上下文地标基于规则化互相关准则进行特征匹配，利用双层匹配矫正策略减少误匹配，得到匹配特征对。提取匹配特征对的中心点作为参考点求解基准图与实时图之间的空间映射关系，进而利用最小二乘拟合一次多项式计算实时图中目标的位置坐标。实验结果表明，针对复杂地面场景，该方法的最大相对定误差不大于3个像素。基本满足成像末制导对自动目标识别算法稳健性好、识别精度高、抗干扰能力强等要求。