大棚智能除草机器人设计

Design of Intelligent Weeding Robot in Greenhouses

针对大棚内除草环境复杂,且杂草种类繁多导致识别困难的特点,设计了一种机械式智能除草机器人。对试验田数据进行采集后,使用Yolov5模型进行150次迭代训练,最终训练出模型平均精度(map50)为82%,为了提高识别准确率,迭代次数增加到300次,最终模型平均精度(map50)为91%,机器人使用了Jetson Orin nano开发板为处理器,Intel D435深度摄像头进行数据采集,图像处理时间为1.2 ms,满足实时处理要求。采用机械式除草,减少了农药使用,该研究可为以后智能除草设备设计提供参考。

Aiming at the characteristics of complex weeding environment in the greenhouse and the difficulty of identification caused by a wide variety of weeds,a mechanical intelligent weeding robot is designed.After collecting the test field data,the Yolov5 model is used for 150 iterations of training,and finally the average accuracy of the model(map50)is 82%,in order to improve the recognition accuracy,the number of iterations is increased to 300,and the average accuracy of the final model(map50)is 91%,and the robot uses Jetson Orin The nano development board is a processor,and the Intel D435 depth camera performs data acquisition,and the image processing time is 1.2 ms,which meets the real-time processing requirements.The use of mechanical weeding reduces the use of pesticides,and this research can provide a reference for the design of intelligent weeding equipment in the future.