基于神经网络和圆谐展开对战斗机俯视图位移、旋转和缩放的识别

神经网络进行畸变不变性模式识别具有普适性、适应性、最优并行性及其在硬件上实现的优点，它可以有效地弥补光学相关模式识别在存在性上的局限性．两者的有效结合对模式识别的发展将起到很大的作用。神经网络和圆谐展开相结合的方式构造复合滤波器可以有效地实现对同一位移、平面旋转和缩放了的图像的畸变不变识别，并给出了该方法用于战斗机俯视图识别的计算机模拟结果。

Scale, Rotation, and Shift Invariant Fighter Plane Planform Recognition Based on Neural Network and Circular Harmonic Expansion

Neural network method possesses universality, adaptability, optimal parallelism and advantages in hardware implementation as well in distortion invariant pattern recognition. This method can offset the limitation of invariant optical correlation pattern recognition efficiently. The method combining neural network with optical correlation pattern recognition will become a promising method in pattern recognition. A method for shift, planar rotation and scale invariant pattern recognition based on BP neural network and circular harmonic expansion was proposed. By using the proposed method, the result of computer simulation on airplane planforms was reported.