摘要
为了实现运煤皮带运行时对进入其中的异物快速准确识别,防止皮带撕裂现象的发生,提出了一种改进的CenterNet运煤皮带异物检测算法。首先,对煤矿井下图像进行预处理,使其适应CenterNet算法,提高网络对目标图像检测的有效性;然后,对网络进行改进,将残差模块中的标准卷积替换成深度可分离卷积,有效降低网络计算量,减少冗余;接着,采用组规范化作为优化规范化方式,降低了对硬件设施的要求;最后,使用加权特征图融合方法,充分利用各层提取的特征,提高网络的检测准确率。实验结果表明,针对异物目标尺寸差异较大且分布不均匀的情况,改进后的CenterNet算法降低了目标的误检率和漏检率,可有效提升检测速度和异物识别精度。
In order to realize rapid and accurate detection of foreign objects entering the coal belt during operation,and prevent belt tearing,an improved CenterNet(ICN)algorithm for coal belt foreign objects of coal belt is proposed in this paper.Firstly,the coal belt images are preprocessed to adapt to the CenterNet algorithm,which can improve the effectiveness of detection.Secondly,the network is improved by replacing the standard convolution in the residual module with the deep separable convolution,which can effectively reduce the network computation and redundancy.Thirdly,the group normalization is adopted as the optimization normalization method,which can reduces the requirements for hardware facilities.Finally,the weighted feature map fusion method is used to make full use of the features extracted from each layer to improve the detection accuracy of the network.The experimental results show that the ICN algorithm reduces the false detection rate and missed detection rate,and can effectively improve the detection speed and accuracy of foreign objects with large size difference and uneven distribution.
作者
任志玲
朱彦存
REN Zhi-ling;ZHU Yan-cun(Faculty of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105,China)
出处
《控制工程》
CSCD
北大核心
2023年第4期703-711,共9页
Control Engineering of China
基金
辽宁省高等学校国(境)外培养项目(2019GJWZD002)
辽宁省高等学校创新团队项目(LT201907)。
关键词
异物检测
CenterNet
组规范化
深度可分离卷积
加权特征融合
Foreign object detection
CenterNet
group normalization
depthwise separable convolution
feature fusionwith weights