IEEE Access, Год журнала: 2023, Номер 11, С. 126398 - 126408

Опубликована: Янв. 1, 2023

As the "Green Concept" gains momentum, state of road infrastructure has emerged as a topic global concern. In parallel, demand for cleaning services provided by sweepers experienced dramatically increase. Consequently, efficiency heavily depends on capabilities sweeper, particularly its capacity environmental perception and ranging. At present, ranging environments is still largely rely artificial observation, inefficient sensors, traditional binocular methods. These conventional techniques fall short in ensuring both cleanliness driving safety sweepers. This study introduces an enhancement to YOLOv8 network, aiming achieve precise integrating predictive frame resolution measurement with stereo vision. Compared method, improve via network effectively avoids inaccuracies misinterpretations stemming from incomplete parallax maps leads heightened levels accuracy safety. Experimental results confirm that enhanced algorithm achieves error rate less than 0.5% under static testing conditions. Furthermore, average can be reduced 0.78% during dynamic scenarios. Our methodology significantly improves precision distance data comparison pre-improvement detection. Post-improvement, model retains portability versatility, making it well-suited permanent integration. demonstrates notable migratability generalisability.

Язык: Английский