多模式特征增强卷积的带式输送机异物检测模型

Foreign object detection model of belt conveyor based on multi-mode feature enhanced convolution

带式输送机异物检测是矿山生产过程中的重要组成部分。针对带式输送机异物特征,笔者提出了一种多模式特征增强卷积神经网络模型,并将其应用于矿山异物检测。该模型使用一种跨级连接Darknet作为模型骨架,以减少图像信息的损失;采用空间特征提取模块,提高模型对模糊物体的特征提取能力;引入注意力融合增强模块,增强相邻特征图之间的信息,实现对多尺度和小目标地准确检测。该方法在带式输送机数据集的准确率达到了93.54%。

Foreign object detection of belt conveyor is an important part in the process of mine production.In view of characteristics of foreign objects in belt conveyors,this paper proposed a multi-mode feature enhanced convolution neural network and applied it to the detection of foreign objects in mines.The algorithm model used a cross-level connection Darknet as the model skeleton to reduce the loss of image information,at the same time,a spatial feature extraction module was used to improve the feature extraction ability of the model for fuzzy objects,and then,a feature fusion enhancement module was introduced to enhance the information between adjacent feature graphs and achieve accurate detection for multi-scale and small targets.This method achieved 93.54%accuracy in the data set of belt conveyors.