基于时空域动态聚合的井下钻场无参考视频质量评价
打开文本图片集
中图分类号：TD67 文献标志码：A
No-reference video quality assessment for underground drilling sites based on spatiotemporal domain dynamic aggregation
WANG Siqian，DONGLihong，YE Ou （College of Artificial Intelligence and Computer Science， Xi'an University of Science and Technology， Xi'an 710054，China）
Abstract：No-Reference Video Quality Assessment （NRVQA）is a key technique for evaluating the video qualityofunderground drilling sites incoal mines and enabling remote monitoring.Existing NRVQA methods are mostlydesigned for general groundscenesandare dificult toachieve satisfactoryperformance in underground drilling environments where composite image distortions are caused bycoal dust and equipment vibration.To addressthis problem，an NRVQA method forunderground drilling sites basedon spatiotemporal domain dynamic aggregation was proposed.Video features of driling site surveillance videos were extracted from two dimensions， namely spatial and motion.The spatial feature extraction branch wasbasedon the Swin Transformer architecture and introduced alocalperception enhancement module to strengthen the representation capabilityof texture and edge details under coal dust interference.The motion feature extraction branch embedded a DeformConv3D deformable convolution module into ResNet to accurately capture the dynamic characteristics of driling rig motion trajectories and coal dust difusion.A spatiotemporal dynamic aggregation module was designed to dynamicallyallcate theweightsofspatialandmotion features，enabling discriminativerepresentationofdifferent distortion types and degrees.The Coal-DB dataset wasconstructed and ablation experimentsandcomparative experiments were conducted.The results showed that the proposed method achieved Spearman rank correlation coefficient， Pearson linearcorrelation coficient， Kendallank correlation coefficient，androot mean square er values of 0.904 3，0.902 3，0.753 6，and 4.684 0，respectively， which were superior to the baseline model and mainstream video quality assessment methods such as VSFA and StableVQA. The predicted video quality scores of this method were closer to the subjective scores.
Key words： underground driling site; video quality assessment; video composite distortion; local perception enhancement; spatiotemporal dynamic aggregation
0引言
近年来，基于计算机视觉的智能监控技术已成为提升煤矿安全生产水平的关键手段，特别是在“人一机-环”全域感知的煤矿智能化建设背景下，视频AI识别技术作为实现安全隐患实时监测与预警的核心支撑，已成为该领域的研究趋势[1]。（剩余13924字）