为了实现绿篱的准确、快速识别,提出了改进的CenterNet绿篱检测算法。在CenterNet的基础上提出了U型连接结构,研究了CenterNet热力图的物理意义并提出了椭圆热力图。最后,改进了原有损失函数以提高预测结果的置信度。实验结果表明:改进...为了实现绿篱的准确、快速识别,提出了改进的CenterNet绿篱检测算法。在CenterNet的基础上提出了U型连接结构,研究了CenterNet热力图的物理意义并提出了椭圆热力图。最后,改进了原有损失函数以提高预测结果的置信度。实验结果表明:改进后的算法对于绿篱在复杂环境中的检测效果有明显提升,改进算法的平均精度(mean average precision, mAP)和F1分数分别为92.92%和83.10%,相比改进前2个指标各自分别平均提升了5.92%和9%。展开更多
The laws of influence of different factors have been analyzed in order to enhance the working efficiency and fatigue life of the cleaning element in brush shape of the sugarcane harvester. Based on the principle of or...The laws of influence of different factors have been analyzed in order to enhance the working efficiency and fatigue life of the cleaning element in brush shape of the sugarcane harvester. Based on the principle of orthogonal experiment design, the virtual-orthogonal-experimental analysis for the cleaning element is carried out on the finite element analysis （FEA） software-ANSYS after analyzing the nonlinear structural behavior in the working procedure. The results are analyzed with the overall balancing method, and then the optimal combination is got, which is made up of different levels of different factors. Also the optimal combination of design parameters of the cleaning element received fiom the virtual experimental analysis is conducted an experiment to confirm that the virtual analysis model and results are right, and the effect of factors on the function of the cleaning element is obtained by more analysis and further optimizing.展开更多
文摘为了实现绿篱的准确、快速识别,提出了改进的CenterNet绿篱检测算法。在CenterNet的基础上提出了U型连接结构,研究了CenterNet热力图的物理意义并提出了椭圆热力图。最后,改进了原有损失函数以提高预测结果的置信度。实验结果表明:改进后的算法对于绿篱在复杂环境中的检测效果有明显提升,改进算法的平均精度(mean average precision, mAP)和F1分数分别为92.92%和83.10%,相比改进前2个指标各自分别平均提升了5.92%和9%。
基金This project is supported by National Natural Science Foundation of China(No.50365001).
文摘The laws of influence of different factors have been analyzed in order to enhance the working efficiency and fatigue life of the cleaning element in brush shape of the sugarcane harvester. Based on the principle of orthogonal experiment design, the virtual-orthogonal-experimental analysis for the cleaning element is carried out on the finite element analysis （FEA） software-ANSYS after analyzing the nonlinear structural behavior in the working procedure. The results are analyzed with the overall balancing method, and then the optimal combination is got, which is made up of different levels of different factors. Also the optimal combination of design parameters of the cleaning element received fiom the virtual experimental analysis is conducted an experiment to confirm that the virtual analysis model and results are right, and the effect of factors on the function of the cleaning element is obtained by more analysis and further optimizing.
