河南地区秸秆焚烧空气污染区域的图像识别算法研究

Image Recognition Algorithm for Air Pollution of Straw Burning in Henan Province

近年来环保部门越来越重视空气污染整治,秸秆焚烧不但产生大量烟雾而且会造成资源浪费,严重污染空气。文中以河南地区秸秆焚烧造成空气污染为研究方向,提出基于有雾图像物理模型的BP神经网络空气污染区域图像识别算法。首先,通过计算入射光衰减项和大气光增强项构建有雾图像的物理模型,利用该模型了解焚烧后有雾污染区域的形成条件,再根据暗原色先验原理对采集图像进行膨胀操作等处理,判断图像是否有雾并对有雾图像建立图像库;其次,以有雾图像库为基础,对焚烧区域中的火焰图像进行特征提取,在考虑干扰因素前提下将目标区域的多个特征参数作为输入神经网络的样本数据,经过神经网络结构训练、学习,预测秸秆焚烧空气污染区域,最终完成对该区域图像的识别过程。实验证明,利用文中BP神经网络图像识别算法实现了对空气污染区域的图像识别。

In recent years, environmental protection departments pay more and more attention to air pollution remediation, straw burning will not only produce large amounts of smoke and will result in waste of resources, serious pollution of the air. Based on the fog image physical model, a BP neural network algorithm for air pollution region image recognition was proposed. The air pollution caused by straw burning in Henan province was studied. First, the physical model of fog image is constructed by computing the attenuation of incident light and the enhancement of atmospheric light. The model is used to understand the formation condition of foggy pollution area after incineration, and then the expansion image is processed according to the a priori principle. Then, based on the fog image database, the flame image in the incineration area is extracted, and the multi-feature parameters of the target area are taken into account in consideration of the interference factors. Then, the fog image is analyzed, As the input data of the neural network, through neural network structure training, learning, and then predict the burning of air pollution areas of straw, and ultimately to complete the region of the image recognition process. Experiments show that the BP neural network image recognition algorithm is used to realize the image recognition of air pollution area.