指针式仪表, 实时目标检测模型, 快速神经网络, 高低频注意力机制, 基于上下文信息的特征融合模块

In the complex environment of crude oil depot, due to the influence of different external interference factors and the limited resources of existing hardware equipment, the accuracy of the model in instrument positioning is low and the computational complexity is high, which is difficult to be popularized and applied.Aiming at this problem, a pointer instrument positioning method for crude oil depot is proposed based on RT-DETR(Real-Time Detection Transformer) network. Firstly, the FasterNet network is introduced to extract the features of partial channels of the input image of the instrument, the parameters and computational complexity of the model are significantly reduced. Secondly, the HiLo attention module is introduced to select the feature of the pointer and scale detail area and the dial smooth area through two paths, which enhances the model's ability to extract the key features of the instrument. Finally, in order to enhance the ability of multi-scale feature fusion and make full use of the feature information of the instrument, the CGFM (Context-Guide Fusion Module) is introduced to further improve the robustness of the model. Experiments show that the detection accuracy of the instrument reaches 97. 6 % , and the parameter quantity of the model is 10. 91 MByte. Compared to the target detection model, it has great advantages.

pointer instrument, real-time detection transformer(RT-DETR), FasterNet, HiLo, context-guide fusion module(CGFM)