摘要
在水位智能识别系统中,采集到的水尺图像可能存在刻度模糊、局部缺失等情况,对水尺识别产生不利影响,针对这一问题提出了一种基于稀疏表示的水位识别方法。该方法利用多幅连续水尺图像对字典进行训练,通过重构残差的比较对样本水尺图像进行分类,根据分类结果计算出水位值。结果表明,该方法对光照变化和局部的遮挡、模糊等具有较强的鲁棒性,可以准确地对水尺兴趣目标图片分类并进行水位计算,计算出的水位与实际水位之间的误差不超过±1 cm。
In the intelligent recognition system of water level,the recognition rate is low due to the water-level ruler image with partial deletion and fuzzy. In order to improve the recognition rate of water level,a novel method which sparse representation based classification was proposed. First,the dictionary was obtained by the number of consecutive water-level ruler images. Then,the water level was acquired through the result of reconstruction residual sample draft classification. The results show that the method is robust to illumination change and partial occlusion,apart from this,it can accurately classify the water level target image and calculate the water level. The calculated error between the water level and the actual water level does not exceed ±1 cm.
出处
《人民黄河》
CAS
北大核心
2016年第12期52-56,共5页
Yellow River
基金
国家"973"计划项目(2013CB328903)
国家自然科学基金资助项目(61403265)
关键词
水位识别
稀疏表示
重构残差
图像识别
water-level recognition
sparse representation
reconstruction residual
image recognition
