摘要
该文以街边车辆识别为背景,针对白日雾霾天气下与夜间路灯强光源下图片成像不清晰问题,采用暗通道先验方式对目标增强,并在YOLOv3算法框架下进行识别。经过实验测试表明,该方法对提高识别的准确率有较大帮助,提高对模糊物体的识别,对较为清晰的实物提高识别概率10%。
Based on vehicle identification as the background, aiming at the problem of unclear image under the day haze weather and the strong light source of street lights at night, the dark channel priori method was used to enhance the target. The recognition was carried out under the framework of YOLOv3 algorithm. The experimental test shows that the method had great help to improve the recognition accuracy, and improved the recognition of fuzzy objects, for relatively clear objective to improve recognition probability of 10%.
作者
冯子博
刘勇
FENG Zi-bo;LIU Yong(Beijing Institute of Graphic Communication,Beijing 102600;AVIC Beijing Precision Engineering Institute for Aircraft Industry,Beijing 100076)
出处
《航空精密制造技术》
2022年第3期6-9,共4页
Aviation Precision Manufacturing Technology
基金
国家自然科学基金资助项目(61773229)
北京市自然科学基金资助项目(KZ201710015010)。