摘要
工矿企业生产过程中不可避免地产生各类污染,其中水污染物排放始终是监控与防治的重要工作。当前工矿企业所采用的传统监控手段如视频或在线设备监控,存在应对突发性水污染事故适应性弱、效率低、成本高、准确性差等问题。针对流动性水体图像连续、动态、全前景、无背景的特点,采用深度学习算法建立动态加权灰度化模型优化彩色图像灰度处理,提出了图像光强信息(灰度值)分析方法;结合时/空间连续图像信息自检方法对水体图像进行在线分析,提出了1种通用水体污染监控方法,开发了基于智能视觉的动态水体污染监控系统,实现了污染状态的高效、准确的定性判断。在工矿企业投用该系统后,运行维护简便,相对于传统人工视频污染监控方式,污染识别准确率提高13%,有效识别率达到99%以上,平均污染识别时间减少3~5 h,实现突发性水污染事故快速响应,大幅降低环保事故发生率;且能够有效降低人员劳动强度,节约企业运营成本。
All kinds of pollutants are inevitably produced in the production process of industrial and mining enterprises,among which the discharge of water pollutants has been always an important work of monitoring and prevention.The traditional monitoring methods adopted by current industrial and mining enterprises,such as video or online equipment monitoring,are often weak in adaptability to the randomness,contingency,and uncertainty of sudden water pollution accidents,and have problems such as low efficiency,high cost,and poor accuracy.Combined with the time/space continuum image information self-check method to analyze the water image online,a universal water pollution monitoring method was proposed,and a dynamic water pollution monitoring system based on intelligent vision was developed,to realize the efficient and accurate qualitative judgment of the pollution state.After the industrial and mining enterprises put the system into use,the operation and maintenance were simple.Compared with the traditional manual video pollution monitoring method,the pollution identification accuracy was increased by 13%,the effective recognition rate was more than 99%,the average pollution identification time was reduced by 3 to 5 hours,the rapid response of sudden water pollution accidents was realized,and the incidence of environmental protection accidents was greatly reduced.This method can effectively reduce the labor intensity of personnel,and save enterprise operating cost.
作者
曾祥吉
鄢锋
李勇刚
潘岩
杨静雅
谭翔天
ZENG Xiangji;YAN Feng;LI Yonggang;PAN Yan;YANG Jingya;TAN Xiangtian(CINF Engineering Co.,Ltd.,Changsha 410019,China;School of Automation,Central South University,Changsha 410083,China)
出处
《环境工程》
CAS
CSCD
北大核心
2023年第11期78-83,122,共7页
Environmental Engineering
基金
国家重点研发计划资助项目“盐湖化工产业集聚区域网络协同制造集成技术研究与应用示范”(2020YFB1713800)
湖南省科技创新计划资助“湖南省科技创新领军人才项目”(2021RC4047)。
关键词
动态水体
动态优化
深度学习
动态加权
图像分析
dynamic water
dynamic optimization
deep learning
dynamic weighting
image analysis
