改进Faster R-CNN算法在箱体盘点中的应用OA

Aiming at the problem of box counting in the dense list of storage shelves,an improved Faster R-CNN target detec-tion algorithm is proposed.Firstly,the EfficientNetb1 is used to replace VGG16 as the backbone feature extraction network,and the MBConv module composed of deep separable convolution and compression and activation network is used to replace the convolution neural network to extract the effective information of the image,to solve the problems of low accuracy and high parameter quantity of the detection network.Secondly,bilinear interpolation method is used to solve the error problem in pooling stage.Finally,in the re-gression prediction stage,the Swish activation function is used to replace the ReLU activation function to reduce the number of ne-crotic neurons.The experimental results show that the mAP value of the improved Faster R-CNN target detection algorithm reaches 66.92%,which is 30.93%higher than the original mAP value,and the model parameter quantity is reduced to 1/5 of the original pa-rameter quantity.In the detection image,the overall number and classification accuracy of box detection have been significantly im-proved,which proves the feasibility of the improved algorithm in the application of box inventory.