基于图像增强技术和深度学习的煤矸石智能识别方法  

Smart Identification Method Based on Image Enhanced Technology and Deep Dearning

作　　者：李锋 LI Feng(China EnergyWuhai Energy Co.,Ltd.,Laoshidan Coal Mine,Wuhai 016000 China)

机构地区：[1]国家能源集团乌海能源有限责任公司老石旦煤矿,内蒙古乌海016000

出　　处：《自动化技术与应用》2025年第3期151-155,共5页Techniques of Automation and Applications

基　　金：国家通源集团乌海能源有限责任公司项目(WHNY-KX-21-05)。

摘　　要：针对煤矸石识别受图像清晰度影响导致效果不佳的问题,提出一种基于图像增强技术和深度学习的煤矸石智能识别方法。首先建立Retinex理论模型,通过快速导向滤波函数对煤矸石图像展开滤波处理,利用Retinex算法完成煤矸石图像的增强处理。其次,将增强后的煤矸石图像输入Snake模型中,通过轮廓点检测获取煤矸石目标区域。最后,将煤矸石目标区域输入MobileNetV3-CBAM网络中通过深度学习完成煤矸石识别。实验结果表明,方法在图像清晰度、目标定位精度、识别率、准确率和识别效率等方面具有良好优势,具有一定应用价值。Aiming at the problem that the recognition of coal gangue is not effective due to the influence of image clarity,an intelligent recognition method of coal gangue based on image enhancement technology and deep learning is proposed.Firstly,the Retinex theoretical model is established,and the coal gangue image is filtered by the fast steering filter function,and the coal gangue image is enhanced by the Retinex algorithm.Secondly,the enhanced coal gangue image is input into the Snake model,and the target area of coal gangue is obtained through contour point detection.Finally,the target area of coal gangue is input into Mobile-NetV3-CBAM network to complete coal gangue identification through deep learning.The experimental results show that the method has good advantages in image clarity,target positioning accuracy,recognition rate,accuracy and recognition efficiency,and has certain application value.

关 键 词：图像增强 深度学习 RETINEX算法 SNAKE模型 智能识别 边缘检测 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]