摘要
针对实际物体的发射率方程难以确定,导致彩色光高温测量过程中建模困难的问题,提出将支持向量机应用于建模过程的方案。研究参数对支持向量机模型的影响,采用改进的网格搜索算法对参数进行优化,利用优化后的最优参数构建支持向量机,得到彩色光高温测量系统的模型。用实验室实测数据对所建模型预测效果进行验证,并与采用神经网络方法建立模型的预测结果进行对比。结果表明,基于支持向量机的仿真模型具有较高的精度和泛化能力。
In view of the fact that object emissivity equation is difficult to determine, which makes it difficult to establish model in colorama high temperature measurement process, hence, support vector machine was employed to solve this problem. By studying the influence of the parameter selection on support vector machine, an improved grid search algo- rithm was used to optimize parameters, and with optimized parameters support vector machine was built and colorama high temperature measurement model was obtained. With the real data measured in laboratory, the predicted effect of the proposed model was verified, and the result was compared with that obtained by neural network, which shows that the model based on support vector machine has higher accuracy and better generalization ability.
出处
《安徽工业大学学报(自然科学版)》
CAS
2015年第4期366-371,共6页
Journal of Anhui University of Technology(Natural Science)
基金
安徽工业大学创新基金(2013074)
关键词
彩色光
高温测量
支持向量机
colorama
high temperature measurement
support vector machine