基于改进YOLOv12的煤矸石智能识别方法

打开文本图片集

中图分类号：TD948.9 文献标志码：A

An intelligent coal gangue recognition method based on improved YOLOvl2

ZHOU Wei2，LI Guangke³

（1.FacultyofEngineering，SIAS University，Zhengzhou451150，China；2.Henan IntelligentManufacturingand DigitalTwinEngineeringResearchCenter，Zhengzhou451150，China；3.SchoolofMechanicalandElectrical Engineering， Zhengzhou University ofLight Industry，Zhengzhou 45ooo1，China）

Abstract：Toaddressthe difficulty ofaccuratelyand eficiently recognizingcoal gangue caused by complex environmental factorssuch as high dustconcentrationandhighly variable iluminationin mines，thisstudy improved the YOLOvl2 network model and proposedan inteligent coal gangue recognitionmethod basedon improved YOLOvl2.A Dual-Scale Sparse Attention （DSSA） mechanism was designed to enhance the model's attention tomulti-scalecoal gangue targetregionsand itsspatial perception capability.AMulti-ConditionFeature Refinement （MCFR）mechanism was designed to perform condition-guided fusion of deep and shalow features， which effectively enhanced the disriminative representation between coal and coal gangue.ADynamic MultiTask Balance Loss（DMTBL）function was constructed to achieve adaptive weight adjustment among localization， clasification，and confidence，thereby strengthening themodel's learningcapability for hardsampleregions. Experimental results showed that the improved YOLOvl2 achieved a precision，recall，and mAP of 96.5% 5 94.9% and 95.8% ，respectively，in the coal gangue recognition task，representing improvements of 3.8% 4.5% ， and 4.5% over the original YOLOvl2，which effectively addressed issues such asmissed detection，false positives，and blurred boundarieswhile maintaining a high inference speed of 47.7 frames per second.Visualization results of activation heatmaps showed that the improved YOLOv12 accurately focused on the target object regions when processing coal gangue with diferent structures and texture complexities， with no obvious background interference，and the activated regions basically cover the main contours of coal blocks and coal gangue.

Key words： coal gangue recognition; improved YOLOvl2； dual-scale sparse attention; multi-condition feature refinement; dynamic multi-task balance loss

0引言

煤矸石是煤炭开采过程中的固体废弃物，如不能妥善处理，不仅会降低煤炭的燃烧效能，其含有的铅、镉、汞、砷、铬等有害重金属还会污染土壤和水源[1-2]。（剩余13618字）