论文部分内容阅读
为提高危险化学品被动红外遥测光谱鉴别正确率,提出应用支持向量机建立鉴别模型。利用野外实测氨气被动红外遥测光谱样本集,变换惩罚因子C对比高斯核函数与多项式核函数的效能,结合网格遍历法搜寻最佳模型参数,建立了基于支持向量机的鉴别模型。基于40个训练样本得到的模型,对包含267个样本的测试样本集的鉴别正确率可达93.6%,明显优于3层网络结构的BP神经网络鉴别模型。实验结果表明,支持向量机鉴别模型是一种有效的危险化学品红外遥测光谱鉴别方法。
In order to improve the accuracy of passive infrared spectroscopy in hazardous chemicals identification, a support vector machine (SVM) is proposed to establish a discriminant model. By using field measured ammonia passive infrared telemetry spectrum sample set, the penalty function C is used to compare the effectiveness of Gaussian kernel function and polynomial kernel function. The grid traversal method is used to search the best model parameters, and a discriminant model based on support vector machine is established. Based on the model obtained from 40 training samples, the recognition accuracy of the test sample set containing 267 samples can reach 93.6%, which is obviously better than the BP neural network identification model of 3-layer network structure. Experimental results show that support vector machine (SVM) discriminant model is an effective method for identifying hazardous chemicals by infrared telemetry.