红外与激光工程, 2018, 47 (8): 0826001, 网络出版: 2018-08-29   

图像能量与对比度的霾检测算法

Haze detection algorithm based on image energy and contrast
作者单位
中国计量大学 计量测试工程学院, 浙江 杭州 310018
摘要
针对目前霾检测方法实时性差且成本较高的问题, 提出一种基于图像能量与对比度的霾检测方法。首先, 对CMOS相机拍摄的图像进行预处理。由于相机受外界影响会出现轻微摆动, 故需对图像进行配准; 其次, 在图像的关键区域中获取目标与水平天空背景的对比度和图像能量两个特征向量; 再次, 将对比度、图像能量、环境湿度作为输入, 将激光粒子计数器测得的实时PM10浓度作为输出, 进行支持向量回归训练, 建立图像和PM10浓度间的关系模型; 最后, 根据得到的模型计算待测图像所对应的PM10浓度。将该方法检测的PM10浓度与激光粒子计数器测得浓度值进行对比, 实验表明该方法检测结果的平均相对误差在10%以内, MSE为0.006 2, 表明预测值与真值拟合程度较好,模型检测的精度较高。在此基础上增加训练样本可进一步提高模型精度。此外, 该方法可针对不同待测环境建立相应的关系模型, 具有较强的灵活性。
Abstract
In view of the poor real-time performance and high cost of haze detection methods, a method based on contrast and image energy was proposed to detect the haze. Firstly, the images taken by the CMOS camera were preprocessed. The image has some slight swing because the camera has been disturbed by the external environment, so the images were registered. Secondly, in the critical region of the image, two contrast vectors of contrast and image energy were obtained. Thirdly, the contrast, image energy and ambient humidity were taken as input, and the real-time PM10 concentration measured by the laser particle counter was used as the output. The relational model between input and output was constructed by training support vector regression(SVR). Finally, the PM10 concentration of the image was calculated using the model. The PM10 concentration detected by this method was compared with that measured by laser particle counter. The average relative error was less than 10% and MSE was 0.006 2, which indicates that the fitting degree between the predicted value and the true value is good and the accuracy of the model was high. On this basis, increasing the training samples can improve the model accuracy. Moreover, the method can establish the corresponding relation model for different environment to be tested, which has strong flexibility.

孔明, 杨天琪, 单良, 郭天太, 王道档, 徐良. 图像能量与对比度的霾检测算法[J]. 红外与激光工程, 2018, 47(8): 0826001. Kong Ming, Yang Tianqi, Shan Liang, Guo Tiantai, Wang Daodang, Xu Liang. Haze detection algorithm based on image energy and contrast[J]. Infrared and Laser Engineering, 2018, 47(8): 0826001.

本文已被 1 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!