通过采用神经网络中的Clipping方法和MonteCarlo修改学习算法，对用于光学模式识别的纯相位二值化匹配滤波器进行了优化设计。计算机模拟结果表明，和传统的纯相位匹配滤波器的相关输出结果相比，其识别输出的信噪比和信号相关峰值得到了明显的提高，从而为今后的光学实现奠定了良好的基础。-Through the use of neural network methods and MonteCarlo modify Clipping learning algorithm for optical pattern recognition of binary phase-only matched filter optimized design. Computer simulation results show that the traditional phase-only matched filter correlation output compared with the results, which identify the output signal to noise ratio and signal correlation peak has been significantly improved, so as to achieve the future optical laid a good foundation.