摘要
目标检测可以替代人眼来进行异常检测,然而目前的目标检测算法存在着在复杂环境下识别速度低,不能实时检测等问题。针对该问题,文中提出了一种基于中心网络(CenterNet)算法,用于复杂环境异常目标检测的方法。CenterNet算法将图像中的检测目标看做一个点,具有计算量小,检测速度快的优点。选择电网数据集进行复杂环境异常目标检测实验。结果表明,该算法的准确率在95%以上,并且基本满足实时性检测需求,与两种其他的检测算法进行对比分析,该方法在检测准确率和检测速度上均优于另两种检测算法。
Object detection can replace human eyes for anomaly detection. However,the current object detection algorithm has the problems of low recognition speed and not real-time detection when facing complex environment. In order to solve this problem,this paper proposes a method based on CenterNet algorithm,which is used to detect the abnormal objects in the complex environment. CenterNet algorithm regards the detected object in the image as a point,which has the advantages of small computation and fast detection speed. The data set of power grid is selected to detect the abnormal objects in the complex site. The results show that the accuracy of the algorithm is over 95%,and it can basically meet the needs of real-time detection. Compared with two other detection algorithms,this method is superior to the other two in recognition accuracy and recognition speed.
作者
赵民
葛云露
丁宁
ZHAO Min;GE Yun-lu;DING Ning(The 22^(th) Research Institute of CETC,Qingdao 266107,China)
出处
《中国电子科学研究院学报》
北大核心
2021年第7期654-660,共7页
Journal of China Academy of Electronics and Information Technology
基金
国家自然科学基金企业创新发展联合基金重点支持项目(U20B2038)。
关键词
目标检测
深度学习
机器学习
人工智能
复杂环境
object detection
deep learning
machine learning
artificial intelligence
complex environment
