煤矿井下钻场打钻是解决瓦斯灾害的重要措施，可以显著提升我国煤矿井下灾害防治水平。为了监测打钻过程并提高打钻效率，需要进行煤矿井下钻场目标检测，即对打钻现场所涉及的重要目标进行识别和定位。相对于传统的煤矿井下钻场目标检测方法，基于深度学习的煤矿井下钻场目标检测方法可以提升目标检测的精度、时效性和稳定性，但需依赖高质量的图像数据集。此外，目标检测结果也易受到光照条件的影响。特别是在井下低光照场景中，由于图像对比度低，目标与背景边界不清晰，导致检测模型较难识别到对应的目标，从而增加了漏检几率。因此，论文针对以上问题提出了相应的解决方案，主要研究内容如下：

针对煤矿井下钻场图像数据资源不足的问题，本文通过采用煤矿用本安型执法记录仪对煤矿井下打钻现场进行拍摄，并经过数据清洗、数据标注、数据检查等步骤，构建了标准化的煤矿井下钻场图像数据集。数据集包含了来自不同钻场和环境背景条件下的70948张图片，涵盖了钻机夹持器、钻机卡盘、煤矿工人、矿井安全帽和钻杆等五类目标，并提供了PASCAL VOC格式的标注文件。此外，通过对现有主流目标检测模型在此数据集上的训练情况进行了对比分析，为相关研究提供了有力的依据和参考。

针对煤矿井下低光照环境引起的图像亮度低、对比度低和细节信息丢失严重的问题，提出一种基于IAT改进的井下低光照图像增强算法。首先，在原网络PEM模块后插入本文提出的多尺度特征融合模块MFFM来弥补多尺度特征融合能力的不足，改善低光区域的细节特征，融合高层语义空间特征信息、低层颜色和纹理信息。其次，引入结合通道注意力和空间注意力优点的通道优先卷积注意力CPCA，抑制图像恢复过程中噪声的生成，放大暗部区域的信息并丰富颜色细节。相较于原有IAT算法，改进后的IAT算法在PSNR和SSIM指标上分别提高了1.19、0.018。相较于其他的低光照图像增强算法HE、SSD、Retinex-Net、KindD、KindD++以及Zero-DCE，改进后的算法在PSNR指标上提高了8.92、4.35、8.72、3.74、2.01、6.17，在SSIM指标上提高了0.126、0.061、0.033、0.022、0.028、0.096。实验结果表明，所提改进算法能够有效提高低光照图像的质量。

针对煤矿井下低光照环境钻场目标检测精度低的问题，提出一种基于YOLOv8改进的井下低光照钻场目标检测算法。首先，利用本文所提出的基于IAT改进的井下低光照图像增强算法作为YOLOv8的预处理模块，用以改善输入图像质量，还原井下低光照图像更多纹理细节。其次，在YOLOv8主干网络中使用提出的SBS模块替换CBS模块，以减少网络下采样时信息的丢失，提高网络对于小目标特征信息的保留能力。同时，在YOLOv8特征融合网络处引入TA轻量级注意力机制，以提升模型对关键特征信息的捕获能力。最后，将CIoU损失函数更换为WIoU损失函数，以提高模型梯度下降速度和收敛速度，进一步提升模型的检测能力。相较于YOLOv5、YOLOv7、YOLOv8、YOLOX以及PPYOLOE，改进后的算法在mAP@0.5指标上提高了1.6%、1.9%、1.5%、8.5%、5.1%，在mAP@0.5:0.95指标上提高了4.5%、6.3%、2.3%、7.2%、8.3%。实验结果表明，所提改进算法能够有效提升井下低光照场景中钻场目标的检测精度。

Coal mine underground drilling is an important measure to solve the gas disaster, can significantly improve our country's coal mine underground disaster prevention level. In order to monitor the drilling process and improve the drilling efficiency, it is necessary to carry out target detection in the coal mine underground drilling site, that is, to identify and locate the important targets involved in the drilling site. Compared with the traditional target detection methods in coal mine drilling sites, the target detection methods in coal mine drilling sites based on deep learning can improve the accuracy, timeliness and stability of target detection, but they rely on high-quality image datasets. In addition, the object detection results are also susceptible to illumination conditions. Especially in the underground low-light scene, due to the low image contrast, the boundary between the target and the background is not clear, which makes it difficult for the detection model to identify the corresponding target, thus increasing the probability of missed detection. Therefore, this paper puts forward the corresponding solutions to the above problems, and the main research contents are as follows:

Aiming at the problem of insufficient image data resources of coal mine drilling site, this paper adopts the intrinsic safety law enforcement recorder for coal mine to photograph the coal mine drilling site, and constructs a standardized image data set of coal mine drilling site after data cleaning, data labeling, data inspection and other steps. The dataset contains 70,948 images from different drilling sites and environmental background conditions, covering five categories of objects such as rig gripper, rig chuck, coal miner, mine safety helmet and drill pipe, and provides annotation files in PASCAL VOC format. In addition, the training of the existing mainstream object detection models on this dataset is compared and analyzed, which provides a strong basis and reference for related research.

Aiming at the problems of low brightness, low contrast and serious loss of detail information of images caused by low light environment in coal mine, an improved underground low light image enhancement algorithm based on IAT is proposed. Firstly, the multi-scale feature fusion module MFFM proposed in this paper is inserted after the original network PEM module to make up for the lack of multi-scale feature fusion ability, improve the detail features of low-light areas, and fuse high-level semantic spatial feature information, low-level color and texture information. Secondly, the channel priority convolutional attention CPCA, which combines the advantages of channel attention and spatial attention, is introduced to suppress the generation of noise in the process of image restoration, enlarge the information of the dark area and enrich the color details. Compared with the original IAT algorithm, the improved IAT algorithm has the PSNR and SSIM indexes increased by 1.19 and 0.018 respectively. Compared with other low-light image enhancement algorithms HE, SSD, Retinex-Net, KindD, KinD++ and Zero-DCE, the improved algorithm has the PSNR index increased by 8.92, 4.35, 8.72, 3.74, 2.01, 6.17. The SSIM index is increased by 0.126, 0.061, 0.033, 0.022, 0.028, 0.096. The experimental results show that the proposed improved algorithm can effectively improve the quality of low-light images.

Aiming at the problem of low accuracy of target detection in the low-light environment of coal mine drilling field, an improved target detection algorithm based on YOLOv8 is proposed. Firstly, the improved underground low-light image enhancement algorithm based on IAT proposed in this paper is used as the preprocessing module of YOLOv8 to improve the input image quality and restore more texture details of underground low-light images. Secondly, the CBS module is replaced by the proposed SBS module in the YOLOv8 backbone network to reduce the loss of information during network downsampling and improve the retention ability of the network for small target feature information. At the same time, the TA lightweight attention mechanism is introduced at the YOLOv8 feature fusion network to improve the capture ability of the model for key feature information. Finally, the CIoU loss function was replaced with the WIoU loss function to improve the gradient descent speed and convergence speed of the model, and further improve the detection ability of the model. Compared with YOLOv5, YOLOv7, YOLOv8, YOLOX and PPYOLOE, the improved algorithm improves the mAP@0.5 index by 1.6%, 1.9%, 1.5%, 8.5% and 5.1%. The index of mAP@0.5:0.95 increased by 4.5%, 6.3%, 2.3%, 7.2%, 8.3%. The experimental results show that the proposed improved algorithm can effectively improve the detection accuracy of drilling site targets in low light scenes.

[1]王双明,申艳军,宋世杰,等.“双碳”目标下煤炭能源地位变化与绿色低碳开发[J].煤炭学报,2023,48(7):2599-2612.

[2]王慧,杨天敏.我国煤炭清洁高效利用现状及发展建议[J].能源,2023(03):64-69.

[3]狄琳锐.新时代煤炭产业转型发展的着力点探析[J].内蒙古煤炭经济,2023(21):150-152.

[4]张建强,宁树正,张莉,等.我国煤炭行业绿色发展现状及实现途径探讨[J].地质论评,2023,69(S1):233-235.

[5]叶彬.打钻与瓦斯抽采精细化管理在煤矿中的应用[J].中国资源综合利用,2020,38(10):63-65.

[6]张小坚.煤矿井下防治水技术与施工研究[J].矿业装备,2023(1):52-53.

[7]蔡波.煤矿地质灾害特征及防治措施研究[J].能源与节能,2023(8):39-41.

[8]管博伦,张立平,朱静波,等.农业病虫害图像数据集构建关键问题及评价方法综述[J].智慧农业(中英文),2023,5(03):17-34.

[9]Xu H, Sun H Z, Wang L B, et al. Urban architectural style recognition and dataset construction method under deep learning of street view images: a case study of Wuhan[J]. ISPRS International Journal of Geo-Information, 2023, 12(7): 264.

[10]Everingham M, Van G L, Williams C K I, et al. The pascal visual object classes (VOC) challenge[J]. International Journal of Computer Vision, 2010, 88(2): 303-338.

[11]Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25.

[12]Lin T Y, Maire M, Belongie S, et al. Microsoft coco: Common objects in context[C]//Computer Vision ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13. Springer International Publishing, 2014: 740-755.

[13]Geiger A, Lenz P, Urtasun R. Are we ready for autonomous driving? the kitti vision benchmark suite[C]//2012 IEEE conference on computer vision and pattern recognition. IEEE, 2012: 3354-3361.

[14]Loh Y P, Chan C S. Getting to know low-light images with the exclusively dark dataset[J]. Computer Vision and Image Understanding, 2019, 178: 30-42.

[15]王丰仪,饶元,罗庆,等.毛桃多模态图像目标检测数据集[J].中国科学数据(中英文网络版),2022,7(04):367-379.

[16]Duan R, Deng H, Tian M, et al. SODA: A large-scale open site object detection dataset for deep learning in construction[J]. Automation in Construction, 2022, 142: 104499.

[17]Wu S, Zhang X, Liu R, et al. A dataset for fire and smoke object detection[J]. Multimedia Tools and Applications, 2023, 82(5): 6707-6726.

[18]Pang Y, Cao J, Li Y, et al. TJU-DHD: A diverse high-resolution dataset for object detection[J]. IEEE Transactions on Image Processing, 2020, 30: 207-219.

[19]李俊杰,李敏,隋正伟,等.中国河南省2016-2021年尾矿库目标检测数据集[J].中国科学数据(中英文网络版),2023,8(04):496-505.

[20]Yang W, Zhang X, Ma B, et al. An open dataset for intelligent recognition and classification of abnormal condition in longwall mining[J]. Scientific Data, 2023, 10(1): 416.

[21]马龙,马腾宇,刘日升.低光照图像增强算法综述[J].中国图象图形学报,2022,27(05):1392-1409.

[22]丁畅,董丽丽,许文海.“直方图”均衡化图像增强技术研究综述[J].计算机工程与应用,2017,53(23):12-17.

[23]Pizer S M, Amburn E P, Austin J D, et al. Adaptive histogram equalization and its variations[J]. Computer vision, graphics, and image processing, 1987, 39(3): 355-368.

[24]王利娟,常霞,任旺.基于加权直方图均衡化彩色图像增强仿真[J].计算机仿真,2021,38(12):126-131.

[25]江巨浪,刘国明,朱柱,等.基于快速模糊聚类的动态多直方图均衡化算法[J].电子学报,2022,50(01):167-176.

[26]文海琼,李建成.基于直方图均衡化的自适应阈值图像增强算法[J].中国集成电路,2022,31(03):38-42.

[27]周辉奎,顾牡丹.自适应加权直方图均衡化的红外图像增强[J].光学技术,2023,49(06):750-755.

[28]Zhuang P, Li C, Wu J. Bayesian retinex underwater image enhancement[J]. Engineering Applications of Artificial Intelligence, 2021, 101: 104171.

[29]张恩齐,孔令胜,郭俊达,等.基于Retinex理论的低照度图像对比度增强算法[J].机电工程技术,2022,51(03):95-98.

[30]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:1-11[2024-04-10].

[31]游达章,陶加涛,张业鹏,等.基于灰度变换及改进Retinex的低照度图像增强[J].红外技术,2023,45(02):161-170.

[32]单永奇,刘文波,滕子煜,等.改进Retinex算法的低照度图像增强研究[J].航空计算技术,2023,53(05):81-84.

[33]何磊,易遵辉,谢永芳,等.基于Retinex先验引导的低光照图像快速增强方法[J/OL].自动化学报:1-12[2024-04-10].

[34]Lore K G, Akintayo A, Sarkar S. LLNet: A deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition, 2017, 61: 650-662.

[35]李正龙,王宏伟,曹文艳,等.基于含噪Retinex模型的煤矿低光照图像增强方法[J].工矿自动化,2023,49(04):70-77.

[36]胡聪,陈绪君,吴雨锴.融合半波注意力机制的低光照图像增强算法研究[J].激光杂志,2024,45(01):109-114.

[37]Cui Z, Li K, Gu L, et al. You only need 90k parameters to adapt light: a light weight transformer for image enhancement and exposure correction[J]. arXiv preprint arXiv:2205.14871, 2022.

[38]黄梦源,常侃,凌铭阳,等.基于层间引导的低光照图像渐进增强算法[J/OL].计算机应用:1-11[2024-04-10].

[39]Chen C L , Peng Y L , Lai-Kuan W . LAU-Net: A low light image enhancer with attention and resizing mechanisms[J]. Signal Processing: Image Communication, 2023, 115.

[40]Viola P, Jones M. Rapid object detection using a boosted cascade of simple features[C]//Proceedings of the 2001 IEEE computer society conference on computer vision and pattern recognition. CVPR 2001. IEEE, 2001, 1: I-I.

[41]Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International journal of computer vision, 2004, 60: 91-110.

[42]Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]//2005 IEEE computer society conference on computer vision and pattern recognition (CVPR'05). Ieee, 2005, 1: 886-893.

[43]Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 580-587.

[44]Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[45]Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28.

[46]He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969.

[47]Liu W, Anguelov D, Erhan D, et al. Ssd: Single shot multibox detector[C]//Computer Vision ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part I 14. Springer International Publishing, 2016: 21-37.

[48]Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[49]Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 7263-7271.

[50]Redmon J, Farhadi A. Yolov3: An incremental improvement[J]. arXiv preprint arXiv:1804.02767, 2018.

[51]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.

[52]Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4: Optimal speed and accuracy of object detection[J]. arXiv preprint arXiv:2004.10934, 2020.

[53]Kim J H, Kim N, Park Y W, et al. Object detection and classification based on YOLO-V5 with improved maritime dataset[J]. Journal of Marine Science and Engineering, 2022, 10(3): 377.

[54]Li C, Li L, Jiang H, et al. YOLOv6: A single-stage object detection framework for industrial applications[J]. arXiv preprint arXiv:2209.02976, 2022.

[55]Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 7464-7475.

[56]Talaat F M, ZainEldin H. An improved fire detection approach based on YOLO-v8 for smart cities[J]. Neural Computing and Applications, 2023, 35(28): 20939-20954.

[57]舒子婷,张泽斌,宋尧哲,等.基于改进YOLOv5的低光照图像目标检测[J].激光与光电子学进展,2023,60(04):77-84.

[58]李昶昱,葛磊.基于YOLOv7的轻量级低照度目标检测算法[J/OL].激光与光电子学进展:1-17[2024-04-11].

[59]陈伟,江志成,田子建,等.基于YOLOv8的煤矿井下人员不安全动作检测算法[J/OL].煤炭科学技术,1-19[2024-04-11].

[60]江泽涛,翟丰硕,钱艺,等.结合特征增强和多尺度感受野的低照度目标检测[J].计算机研究与发展,2023,60(04):903-915.

[61]Carion N, Massa F, Synnaeve G, et al. End-to-end object detection with transformers[C]//European conference on computer vision. Cham: Springer International Publishing, 2020: 213-229.

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

[63]程德强,徐进洋,寇旗旗,等.融合残差信息轻量级网络的运煤皮带异物分类[J].煤炭学报,2022,47(03):1361-1369.

[64]程德强,陈杰,寇旗旗,等.融合层次特征和注意力机制的轻量化矿井图像超分辨率重建方法[J].仪器仪表学报,2022,43(08):73-84.

[65]程德强,钱建生,郭星歌,等.煤矿安全生产视频AI识别关键技术研究综述[J].煤炭科学技术,2023,51(02):349-365.

[66]Huang H, Chen Z, Zou Y, et al. Channel prior convolutional attention for medical image segmentation[J]. arxiv preprint arxiv:2306.05196, 2023.

[67]Hao S, Han X, Guo Y, et al. Low-light image enhancement with semi-decoupled decomposition[J]. IEEE transactions on multimedia, 2020, 22(12): 3025-3038.

[68]Wei C, Wang W, Yang W, et al. Deep retinex decomposition for low-light enhancement[J]. arXiv preprint arXiv:1808.04560, 2018.

[69]Zhang Y, Zhang J, Guo X. Kindling the darkness: A practical low-light image enhancer[C]//Proceedings of the 27th ACM international conference on multimedia. 2019: 1632-1640.

[70]Zhang Y, Guo X, Ma J, et al. Beyond brightening low-light images[J]. International Journal of Computer Vision, 2021, 129: 1013-1037.

[71]Guo C, Li C, Guo J, et al. Zero-reference deep curve estimation for low-light image enhancement[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 1780-1789.

[72]Sunkara R, Luo T. No more strided convolutions or pooling: A new CNN building block for low-resolution images and small objects[C]//Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2022: 443-459.

[73]Misra D, Nalamada T, Arasanipalai A U, et al. Rotate to attend: Convolutional triplet attention module[C]//Proceedings of the IEEE/CVF winter conference on applications of computer vision. 2021: 3139-3148.

[74]Ge Z, Liu S, Wang F, et al. Yolox: Exceeding yolo series in 2021[J]. arXiv preprint arXiv:2107.08430, 2021.

[75]Xu S, Wang X, Lv W, et al. PP-YOLOE: An evolved version of YOLO[J]. arXiv preprint arXiv:2203.16250, 2022.

[76]Selvaraju R R, Cogswell M, Das A, et al. Grad-cam: Visual explanations from deep networks via gradient-based localization[C]//Proceedings of the IEEE international conference on computer vision. 2017: 618-626.