改进的YOLOV3算法在小目标检测中的研究与应用

The Research and Application of Small Target Detection Based on Improved YOLOV3 Algorithm

战场目标的图像检测与识别对于战场监视、侦察、毁伤状态评估和火控系统研究等具有重要作用。以坦克装甲目标为研究对象,选用识别精度高、速度快的YOLOV3为基础目标检测模型,针对复杂战场环境中获取图像目标特征信息少的问题,引入多尺度特征增强结构的方法对YOLOV3模型进行改进,通过丰富特征图多样性的方式,提高模型性能。在坦克数据集上的实验结果表明,改进后的算法对于复杂战场环境下的小目标特征具有更强的敏感性,较大程度上增强了模型的识别精度。

The image detection and identification of battlefield targets have played an important role in battlefield surveillance,reconnaissance,damage assessment and fire control system research,etc.Based on tank armored vehicle as the research target,the high accuracy and high speed YOLOV3 is selected as the basic target detection model,the contents of complex images for target feature information are few,the method of multi-scale features enhanced structure is introduced to improve YOLOV3 model improve the model performance and to with rich features figure diversity.The experimental results on the tank dataset show that the improved algorithm is more sensitive to the features of small targets in a complex background,which greatly enhances the detection and recognition accuracy of the model.