摘要
最小可辨温差(MRTD)是评价红外热像仪的重要技术指标,传统的测量方法都是基于测量人员的主观判读得到的,结果重复性差。为了解决这一问题,提出基于人工神经网络的客观评测方法,提出两种特征量——杆均值与背景极值对比度和相邻极值差与均值对比度,使用数字CCD摄像机代替人眼,采集大量图像数据输入人工神经网络,使用BP网络的LM算法进行训练。训练好的神经网络具有类似人的视觉系统的判读能力,可以由计算机代替人工对不同频率、不同温差下的4杆靶图像进行判读。通过Matlab模型和软件验证了系统的可用性,大量测试结果表明:测量结果准确,具有很好的重复性,与相同测量环境条件下由人眼主观测量的MRTD结果相吻合。采用该技术研制的测试设备已成功地应用于某试验单位的检测中,大大提高了测量效率和测量结果的准确性。
Minimum Resolvable Temperature Difference（MRTD） is an important technical parameter to estimate the infrared thermal imager.The traditional measurement method based on the testers′ subjective judging has poor reproducibility.To solve this problem,an objective evaluation method based on Artificial Neural Network（ANN） was developed.Two image features were put forward in the paper.Instead of human eyes,large numbers of image data were acquired by digital CCD and transferred to ANN.Then the ANN was trained by BP-Neural-Networks-based LM algorithm.The judging ability of well trained ANN was similar to the visual system of human eyes.The feasibility of the proposed method was proved through Matlab model and the developed software.And the test results show good accuracy and repeatability of the proposed method.It agrees well with the subjective one under the same test environment.The software system and equipment based on the proposed method had been successfully used in some test cases and greatly improved the test efficiency and accuracy.
出处
《红外与激光工程》
EI
CSCD
北大核心
2010年第4期611-613,654,共4页
Infrared and Laser Engineering
基金
预研项目
关键词
MRTD客观评测
红外测量
人工神经网络
红外热像仪
Objective evaluation for MRTD
Infrared measurement
Artificial neural network
Infrared thermal imaging system
