红外序列图像的主成分分析算法研究

Research on principal component analysis algorithm for infrared image sequence

采用主动式红外热成像检测材料缺陷,在控制外界环境影响因素的前提下,实验研究并分析了热激励时间和探测距离两个检测因素的影响。同步采集得到最佳参数下玻璃钢平底孔试块的红外热图像序列。主动式红外无损检测通常获得包含数百帧图的红外序列,为了获取反映整个图像序列的主要信息,采用了主成分分析算法(principal component analysis,PCA)对图像序列进行处理。首次探究了融合区间对PCA算法处理结果的影响,提出了可依据温差峰值下降的百分数来选择融合区间,并对处理结果进行了主客观的对比评价分析。研究结果表明当温差峰值下降到80%时,选择大于该值对应的序列图像帧数作为融合区间时,PCA处理的效果最佳。最后探讨了PCA处理中减少红外镜头反光等影响因素的策略。

In this paper,active infrared thermography is used to detect material defects.Under the premise of controlling the influence of the external environment,the effects of two detection factors,thermal excitation time and the detection distance,are experimentally studied and analyzed.The infrared image sequence of the glass fiber reinforced plastic test block with flat bottom holes under the best parameters is acquired synchronously.The infrared image sequence usually contains hundreds or thousands of frames of images.In order to obtain the main information representing the entire image sequence,the principal component analysis algorithm(PCA)is used to process the image sequence.The influence of the fusion interval on the processing results of the PCA algorithm is explored for the first time,and the fusion interval can be selected according to the percentage of the temperature peak drop.The objective and subjective comparative evaluation and analysis of the processing results are also made.The experimental results show that when the peak temperature difference drops to 80%,selecting the sequence image frame number greater than this value as the fusion interval,the PCA processing has the best effect.Finally,the strategy is discussed to reduce the influence factors,such as the reflection of the infrared lens,in PCA processing.