基于图像识别的液位在线读取  

Liquid level on-line reading based on image recognition

作　　者：代素梅[1] 张贵涛[1] 钟旭[1] 宋冲[1] 贾瑞清[1] 

机构地区：[1]中国矿业大学(北京)机电与信息工程学院,北京100083

出　　处：《辽宁工程技术大学学报（自然科学版）》2009年第2期284-287,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金项目资助(50475094)

摘　　要：为实现液位实验结果在远程实验中的在线读取,提出应用图像处理的方法进行液位识别。经过背景差分、二值化和中值滤波的图像处理过程,建立了液位读取计算模型,把图像待识别部分和特征匹配模板都用一维向量表示是一种较好的建立计算模型的方法。同时研究液位误差的影响因素,摄像头光轴与玻璃管的垂直度和玻璃管与水平面的垂直度是引起误差的关键;确定了误差分析模型并用于校正识别结果。实验结果表明,液位图像识别的精度较高,可满足实际应用要求,是一种具有很好发展前景的液位读取方法。In order to read liquid level on-line during remote experiment, it is proposed that applying image processing is used to recognize the liquid level. Through the process of image background difference, image binarization and median filter, the liquid level recognition model was established. In the recognition model, both the parts to be recognized and the feature match template were shown by one dimension vectors. Meanwhile, factors that influence liquid level error were also studied. The key factors that cause errors are verticality between camera optic axis and glass tubes, and verticality between glass tubes and the horizontal. Hence the model of error analysis was drawn, and was used for adjusting the results. Experimental results indicate that the identification accuracy is good enough for practical use. Image processing is a new method for identifying liquid level and has a good development prospect.

关 键 词：液位 图像处理 特征匹配 误差分析 

分 类 号：TH-39[机械工程]