摘要
水华问题是全球内陆湖普遍存在的问题,不但会造成大量的经济赤字,而且威胁到了人类的生存环境。传统的治理方式比如人工法、化学法或生物法等都依赖于明确蓝藻的爆发情况和重灾区域,针对这种情况,提出用深度神经网络模型实现基于图像的蓝藻的语义分割,用于定位蓝藻爆发的区域。实验过程中发现复杂光照条件对分割结果影响非常大的问题,故设计一种基于HIS图像格式的光照处理算法。借助I通道可以分离的特性,将图像的光强照度指数归一化到合理的光照区间内。实验证明该算法有效地提高了复杂光照条件下的分割准确率。
The issue of water bloom is a ubiquitous problem of the inland lakes in the world.It not only causes a lot of economic deficits,but also threatens the living environment of mankind.Traditional methods of treatment such as artificial methods,chemical methods or biological methods rely on the clear outbreak of cyanobacteria and hardest hit areas.In view of this situation,it was proposed that the deep neural network model was used to realize semantic segmentation of image-based cyanobacteria for the purpose of locating the area where cyanobacteria breaks out.During the experiment,it was found that the complex lighting conditions had a very big impact on the segmentation results.Therefore,a light processing algorithm based on HIS image format was designed.With the help of the characteristics that I channel can be separated,the illumination intensity index of image was normalized to reasonable of the light range.Experimental results showed that the proposed algorithm effectively improved the segmentation accuracy under complex lighting conditions.
作者
罗艾娜
郑骏
Luo Aina;Zheng Jun(East China Normal University,Shanghai 200062,China)
出处
《计算机应用与软件》
北大核心
2018年第4期254-259,共6页
Computer Applications and Software
作者简介
罗艾娜,硕士生,主研领域:图像图形处理与计算机视觉。;郑骏,教授。
