基于STM 32F103单片机的远程无线瓦斯浓度监测系统

In view of the shortcomings of traditional gas concentration monitoring system, a remote wireless gas concentration moni- toring system based on STM32F103 is designed. The system uses STM32F103 microcontroller to control the sensor to realize the ac- quisition and processing of the underground gas concentration in real time. When the gas concentration exceeds the set upper limit, the system will send the sound and light alarm, transmit the collected data to the GPRS DTU, and then transmit the generated alarm signal and collected data to the monitoring center in time by using TCP/IP. The system can realize real-time monitoring by SMS or telephone, which greatly reduces the waste of human and financial resources.     

基于YOLOv7的交通目标检测算法研究

针对交通场景中由光照、遮挡、目标小以及背景复杂等因素导致目标检测精度低,易出现漏检和误检问题的情况,提出了一种基于YOLOv7的交通目标检测算法;该算法在主干网络中融入多头注意力机制,以增强网络特征学习能力,从而更好地捕获数据和特征内部的相关性;在YOLOv7颈部网络引入协调注意力模块(CA),将位置信息嵌入到注意力机制中,忽略无关信息的干扰,以增强网络的特征提取能力;增加一个多尺度检测网络,以增强模型对不同尺度目标的检测能力;将CIoU损失函数更改为SIoU函数,以减少模型收敛不稳定问题,提高模型的鲁棒性;实验结果表明,改进的算法在BDD100K公开数据集上的检测精度和速度分别达到了59.8% mAP和96.2 FPS,相比原算法检测精度提高了2.5%;这表明改进的算法在满足实时性要求的同时,具备良好的检测精度,适用于复杂情况下的交通目标检测任务。