在煤矿生产过程中，煤中混杂 的异物不仅对煤炭品质产生不利影响，还可能对生产设备和生态环境造成破坏。因此，快速而准确地识别出煤中的异物至关重要。然而煤矿井下监控图像常常光照暗且不均匀，煤中异物识别效率有待进一步提高，因此本文提出一种改善煤矿井下图像质量的增强方法，并提出一种高效的煤中异物识别方法。

（1）针对矿井图像照度低和光照不均问题，提出一种改进的基于图像结构分解的矿井图像增强方法。该方法以变分优化Retinex为基础理论，充分利用图像的结构纹理特性构建变分优化函数，在目标函数的约束下依次求解出图像的结构分量、纹理分量、噪声分量，并进行分量重构，在减少了图像噪声的同时，获得了纹理清晰的初始亮度分量。为避免图像亮度的过增强，改进权重分布的伽马校正算法处理初始亮度分量以获得最终的亮度分量。最后将图像转化到RGB颜色空间，得到增强图像。实验仿真表明，该方法处理后的图像能保证图像边缘纹理清晰，同时提高了亮度和对比度，减少了图像增强过程中的光晕伪影。与近年的图像增强方法相比，本文方法增强后图像的视觉效果更好，在客观评价指标上整体表现更优，对于煤矿井下图像具有较好的适用性。

（2） 针对煤中混入异物的识别问题，本文提出一种改进的YOLOv7检测方法（Improved -YOLOv7）。该方法从三个方面对原YOLOv7进行改进。首先使用改进后的金字塔视觉Transformer重新设计了YOLOv7的主干特征提取网络，使网络更适应于矿井中的图像识别任务。其次，在颈部网络设计轻量化特征融合卷积块和下采样模块，并融合坐标注意力提升模型特征图的多样性和表达能力。最后将YOLOv7的定位损失改进为SIOU损失函数，进一步提升模型精度。实验结果表明，在非煤异物检测任务中，Improved-YOLOv7模型相较于近年来先进的目标检测算法，检测精度得到明显提升，mAP可达96.83%，相较于原YOLOv7-tiny的mAP提升了6.21%，并且满足实时性要求，检测速率达24fps，最后对算法进行了可行性分析，便于在实际环境中部署和应用。

During the coal mining process, the presence of foreign objects mixed in coal not only adversely affects coal quality but also poses potential damage to production equipment and the ecological environment. Therefore, it is crucial to identify foreign objects in coal rapidly and accurately. However, the monitoring images of underground coal mine are often dark and uneven, and the identification efficiency of foreign objects in coal needs to be further improved. Therefore, an enhancement method to improve the image quality of underground coal mine and an efficient identification method of foreign bodies in coal are proposed in this thesis.

(1)An image enhancement method based on image structure decomposition is proposed to solve the problems of low illumination and uneven lighting in mine images.Based on the theory of variational optimization Retinex, this method makes full use of the structural texture characteristics of the image to construct the variational optimization function. Under the constraint of object function, the structural component, texture component and noise component of the image are solved in turn,then the components are reconstructed and the initial brightness component with clear texture is obtained. To avoid excessive enhancement of image brightness, an improved gamma correction algorithm is employed to process the initial brightness component, resulting in the final brightness component. Finally, the image is transformed into the RGB color space to obtain the enhanced image. Experimental simulations demonstrate that the processed images using this method ensure clear edge textures, improve brightness and contrast, and reduce halo artifacts during the image enhancement process. Compared with recent image enhancement methods, the proposed method exhibits better visual effects and overall performance on objective evaluation metrics, making it highly applicable to underground coal mine images.

(2)Addressing the issue of identifying foreign objects mixed within coal, an Improved-YOLOv7 detection method is proposed in this thesis. This method enhances the original YOLOv7 detection method in three aspects. Firstly, the improved pyramid visual Transformer was used to redesign the backbone feature extraction network of YOLOv7 to make the network more adapted to image recognition tasks in the mine. Secondly, lightweight feature fusion convolution blocks and downsampling modules are designed in the neck network, along with the fusion of coordinate attention to enhance the diversity and expressive power of model feature maps. Finally, the localization loss of YOLOv7 is improved to SIOU loss function, further enhancing the model's accuracy. Experimental results demonstrate that in the task of detecting non-coal impurities, the Improved-YOLOv7 model achieves a significantly improved detection accuracy compared to recent advanced object detection algorithms, the mean average precision (mAP) is 96.83%. Compared with the original YOLOv7-tiny, the mAP is improved by 6.21%, and it meets real-time requirements with a detection rate of 24fps. Finally, a feasibility analysis was conducted on the algorithm to facilitate deployment and application in practical environments

[1]丁恩杰,俞啸,夏冰,等.矿山信息化发展及以数字孪生为核心的智慧矿山关键技术[J]. 煤炭学报, 2022,47(01): 564-578.

[2]曹虎奇.煤矿带式输送机撕带断带研究分析[J]. 煤炭科学技术, 2015, 43(S2): 130-134.

[3]张强,张润鑫.煤矿智能化开采煤岩识别技术综述[J]. 煤炭科学技术,2022, 50(02): 1-26.

[4]李济军.以自动化信息化融合为基础的智慧矿山建设探析[J]. 当代化工研究, 2023,(16):194-196.

[5]彭大鑫,甄彤,李智慧.低光照图像增强研究方法综述[J]. 计算机工程与应用,2023,59(18):14-27.

[6]Jobson D J ,Rahman Z,Woodell G A. Properties and performance of a center/surround retinex[J]. IEEE Transactions on Image Processing, 1997, 6(3): 451-462.

[7]Rahman Z, Jobson D J, Woodell G A. Multi-scale retinex for color image enhancement[C]. 3rd IEEE International Conference on Image Processing . Lausanne,1996 :1003-1006.

[8]Jobson D J , Rahman Z, Woodell G A, A multiscale retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing,1997,7(6) :965-976.

[9]张立亚,郝博南,孟庆勇,等. 基于HSV空间改进融合Retinex算法的井下图像增强方法[J]. 煤炭学报, 2020, 45(S1): 532-540.

[10]唐守锋,史可,仝光明,等.一种矿井低照度图像增强算法[J]. 工矿自动化, 2021, 47(10):32-36.

[11]洪炎,朱丹萍,龚平顺.基于TopHat加权引导滤波的Retinex矿井图像增强算法[J]. 工矿自动化, 2022, 48(8): 43-49.

[12]苏波,李超,王莉.基于多权重融合策略的Retinex矿井图像增强算法[J/OL]. 煤炭学报(网络首发版), 2022, 1-11.

[13]张红,索霆锋,宋婉莹.基于结构纹理分解的矿井图像增强方法[J]. 工矿自动化, 2024, 50(3)：56-64.

[14]Fu X, Zeng D, Huang Y, et al. A weighted variational model for simultaneous reflectance and illumination estimation[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2016: 2782-2890.

[15]谢凤英,汤萌,张蕊.基于Retinex的图像增强方法综述[J].数据采集与处理, 2019, 34(01): 1-11.

[16]武亚红.不均匀低照度低质图像增强算法研究[D].南京邮电大学,2021.

[17]Cai B, Xu X, Guo K, et al. A joint intrinsic-extrinsic prior model for retinex[C]. IEEE International Conference on Computer Vision(ICCV), 2017:4020-4029.

[18]Hao S, Han X, Guo Y, et al. Lowlight image enhancement with semi-decoupled decomposition[J]. IEEE Transactions on Multimedia, 2020, 22(12): 3025-3038.

[19]Xu J, HOU Y, RRN D, et al. STAR: A structure and texture aware retinex model[J]. IEEE Transactions on Image Processing, 2020, 29: 5022-5037.

[20]郭永存,何磊,刘普壮,等. 煤矸双能 X 射线图像多维度分析识别方法[J]. 煤炭学报,

2021, 46(01): 300-309.

[21]宋亮, 刘善军, 虞茉莉,等. 基于可见-近红外和热红外光谱联合分析的煤和矸石分类方法研究[J]. 光谱学与光谱分析, 2017, 37(02): 416-422.

[22]Zhang H, Zhang Y. An Improved Sparrow Search Algorithm for Optimizing Support Vector Machines[J]. IEEE Access, 2023, 11: 8199-8206.

[23]李曼,段雍,曹现刚,等.煤矸分选机器人图像识别方法和系统[J]. 煤炭学报, 2020, 45(10): 3636-3644.

[24]史凌凯,耿毅德,王宏伟,等.基于改进Mask R-CNN的刮板输送机铁质异物多目标检测[J]. 工矿自动化, 2022, 48(10): 55-61.

[25]吴守鹏,丁恩杰,俞啸. 基于改进 FPN 的输送带异物识别方法[J].煤矿安全，2019, 50(12): 127-130.

[26]Yuanbin W, Yujing W, Langfei D. Video detection of foreign objects on the surface of belt conveyor underground coal mine based on improved SSD[J]. Journal of Ambient Intelligence and Humanized Computing, 2020, 1-10.

[27]任志玲,朱彦存.改进CenterNet算法的煤矿皮带运输异物识别研究[J].控制工程, 2023, 30(04): 703-711.

[28]胡璟皓,高妍,张红娟,等.基于深度学习的带式输送机非煤异物识别方法[J]. 工矿自动化, 2021, 47(06): 57-62.

[29]Mengchao Z, Yueshuai C, Kai J, et al. Proactive measures to prevent conveyor belt Failures: Deep Learning-based faster foreign object detection[J]. Engineering Failure Analysis, 2022,141:106653.

[30]Sharifani K, Amini M. Machine Learning and Deep Learning: A Review of Methods and Applications[J]. World Information Technology and Engineering Journal, 2023, 10(07): 3897-3904.

[31]Qinghua M, Shikun L, Xin H, et al. Coal Mine Belt Conveyor Foreign Objects Recognition Method of Improved YOLOv5 Algorithm with Defogging and Deblurring[J].Energies (Basel), 2022, 15(24):9504

[32]樊红卫,刘金鹏,曹现刚,等.低照度尘雾下煤、异物及输送带早期损伤多尺度目标智能检测方法[J/OL]. 煤炭学报(网络首发版), 2022,1-12.

[33]张释如,黄综浏,张袁浩,等.基于改进YOLOv5的煤矸识别研究[J]. 工矿自动化,2022, 48(11):39-44.

[34]唐俊,李敬兆,石晴,等.基于Faster-YOLOv7的带式输送机异物实时检测[J]. 工矿自动化,2023, 49(11):46-52+66.

[35]Chen K, Du B, Wang Y, et al. The real-time detection method for coal gangue based on YOLOv8s-GSC[J]. Journal of Real-Time Image Processing, 2024, 21(2): 1-12.

[36]Li N, Ma L, Yu G, et al. Survey on evolutionary deep learning: Principles, algorithms, applications, and open issues[J]. ACM Computing Surveys, 2023, 56(2): 1-34.

[37]邵延华,张铎,楚红雨,等.基于深度学习的YOLO目标检测综述[J]. 电子与信息学报,2022, 44(10):3697-3708.

[38]Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.

[39]Tumas P, Serackis A. Automated image annotation based on YOLOv3[C]. 2018 IEEE 6th Workshop on Advances in Information, Electronic and Electrical Engineering (AIEEE), 2018: 1-3.

[40]Wang CY, Bochkovskiy A, Liao H Y M. Scaled-YOLOv4: Scaling cross stage partial network[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2021: 13029-13038.

[41]Li N, Wang M, Yang G, et al. DENS-YOLOv6: A small object detection model for garbage detection on water surface[J]. Multimedia Tools and Applications, 2023: 1-21.

[42]Wang CY, Bochkovskiy A, Liao HYM. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C].IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2023: 7464-7475.

[43]Talaat Fm, ZainEldin H. An improved fire detection approach based on YOLOv8 for smart cities[J]. Neural Computing and Applications, 2023, 35(28): 2039-2095.

[44]刘颖,刘红燕,范九伦,等.基于深度学习的小目标检测研究与应用综述[J]. 电子学报,2020, 48(03): 590-601.

[45]张欣.基于YOLO算法的遥感图像飞机目标检测技术研究[D]. 中国科学院大学(中国科学院长春光学精密机械与物理研究所), 2021.

[46]Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.

[47]Kocoń J, Cichecki I, Kaszyca O, et al. ChatGPT: Jack of all trades, master of none[J]. Information Fusion, 2023: 101861.

[48]刘文婷,卢新明.基于计算机视觉的Transformer研究进展[J]. 计算机工程与应用,2022, 58(06): 1-16.

[49]Chien C, Kinoshita Y, Kiya H. A Noise-aware Enhancement Method for Underexposed Images[C]. IEEE International Conference on Consumer Electronics Asia (ICCE-Asia), 2019: 131-134.

[50]LI Z, ZHENG J, ZHU Z, et.al. Weighted guided image filtering[J]. IEEE Transactions on Image Processing, 2014, 24(1): 120-129.

[51]Li M, Liu J, Yang W, et al. Structure-revealing low-light image enhancement via robust retinex model[J]. IEEE Transactions on Image Processing, 2018, 27(6): 2828-2841.

[52]HR Sheikh, AC Bovik Image information and visual quality[J]. IEEE Transactions on Image Processing, 2006, 15(2): 430-444.

[53]A Mittal, R Soundararajan Bovik A C. Making a “completely blind” image quality analyzer[J]. IEEE Signal Processing Letters, 2012, 20(3): 209-212.

[54]Wang W, Xie E, Li X, et al. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions[C]. IEEE International Conference on Computer Vision(ICCV) ,2021: 568-578.

[55]Wang W, Xie E, Li X, et al. Pvt v2: Improved baselines with pyramid vision transformer[J]. Computational Visual Media, 2022, 8(3): 415-424.

[56]Han K, Wang Y, Tian Q, et al. GhostNet:More features from cheap operations[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2020: 1580-1589.

[57]Zhao X, Song Y. Improved Ship Detection with YOLOv8 Enhanced with MobileViT and GSConv[J]. Electronics, 2023, 12(22): 4666.

[58]王剑哲,吴秦.坐标注意力特征金字塔的显著性目标检测算法[J]. 计算机科学与探索,2023, 17(01):154-165.

[59]Du S, Zhang B, Zhang P. Scale-sensitive IOU loss: An improved regression loss function in remote sensing object detection[J]. IEEE Access, 2021,9: 14128-14122.