摘要
提出一种先估计姿态后识别目标的新方法,克服了传统三维目标识别算法的多义性问题。首先根据航迹三次样条插值函数估算目标姿态,避免了复杂的图像处理过程,计算速度快;然后在虚拟环境中将候选目标调整到该姿态并生成相应的二维模型;最后根据不变矩特征的欧氏距离识别出目标。实验表明,该方法具有较好的准确性和实时性,对于提高三维目标识别的效率具有理论意义。
A new method was proposed to recognize target after attitude estimation, which overcame the ambiguity problem of traditional 3D target recognition algorithms. Firstly, the attitude of the target was estimated according to cubic spline interpolation functions of the track, which avoided complex image processing and had higher speed. Then, the target candidates were adjusted to this attitude in virtual environment and the corresponding 2D models were generated. At last, the target was recognized according to Euclid distance of moment invariants. The experiment showed that this method has higher precision and better real-time performance, which is theoretically significant for improving the efficiency of 3D targetrecognition.
出处
《电光与控制》
北大核心
2011年第2期45-48,共4页
Electronics Optics & Control
基金
武器装备军内科研项目
关键词
三维目标识别
姿态角
三次样条插值
不变矩
模型匹配
3D target recognition
attitude angle
cubic spline interpolation
moment invariant
modelmatching
