一种基于多核学习的特征融合方法

• Format: Patent
• Language: Chinese
• Published: 15.08.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

本发明属于多模复合制导信息融合技术领域，公开了一种基于多核学习的特征融合方法，具体为：1)对雷达、红外训练特征集分别进行标准化处理；2)用神经网络f、g分别抽取雷达、红外的非线性特征X(f)、Y(g)；3)在神经网络的输出层构造融合准则函数，使得相关系数corr(X(f),Y(g))最大化；4)优化神经网络得到融合后的特征向量；5)确定预选取的基核函数；6)采用加权求和的方式得到合成核；7)利用简单多核学习算法训练合成核；8)在线鉴别时，对雷达和红外的测试特征集合重复步骤1～4，用训练完毕的合成核替代传统支持向量机中的单一核函数；9)确认待攻击目标的身份；在保证类间散布矩阵最大，类内散步矩阵最小的同时，降低特征维度，提高了信息融合的鉴别性能。 The invention belongs to the technical field of multimode composite guidance information fusion, and discloses a feature fusion method based on multi-kernel learning, which specifically comprises thefollowing steps: 1) performing standardization processing on radar and infrared training feature sets respectively; 2) respectively extracting nonlinear features X (f) and Y (g) of radar and infraredby using neural networks f and g; 3) constructing a fusion criterion function on an output layer of the neural network to maximize a correlation coefficient corr (X (f), Y (g)); 4) optimizing the neural network to obtain a fused feature vector; 5) determining a pre-selected basis kernel function; 6)