A simulation approach based on a full-wave numerical method is presented to study electromagnetic characteristics by complex targets. How to validate simulation results is considered thoroughly under no analytical and...A simulation approach based on a full-wave numerical method is presented to study electromagnetic characteristics by complex targets. How to validate simulation results is considered thoroughly under no analytical and measured data, where a double-check criterion is designed for our simulation approach. As an example, the scattering of F-117A is studied by using our simulation approach under all polarizations, different frequency bands, incident and scattering directions, etc., some of which, such as cross-polarization, bistatic RCS, have not been considered in the previous literature.展开更多
The electromagnetic scattering computation has developed rapidly for many years; some computing problems for complex and coated targets cannot be solved by using the existing theory and computing models. A computing m...The electromagnetic scattering computation has developed rapidly for many years; some computing problems for complex and coated targets cannot be solved by using the existing theory and computing models. A computing model based on data is established for making up the insufficiency of theoretic models. Based on the ＂support vector regression method＂, which is formulated on the principle of minimizing a structural risk, a data model to predicate the unknown radar cross section of some appointed targets is given. Comparison between the actual data and the results of this predicting model based on support vector regression method proved that the support vector regression method is workable and with a comparative precision.展开更多
A practical approach for calculating the RCS (Radar Cross Section) of complex targets modeled with wire-grid-frame is presented. A way for generating a polyhedron model (facet-wedge model) with the wire-grid-frame dat...A practical approach for calculating the RCS (Radar Cross Section) of complex targets modeled with wire-grid-frame is presented. A way for generating a polyhedron model (facet-wedge model) with the wire-grid-frame data is described. For storing and reading the data of the polyhedron model in an easy way, a data structure is given.展开更多
针对形状不规则复杂面目标多弹瞄准点优化算法计算效率低、稳定性差、优化能力不足的问题,提出一种基于弹药圆概率偏差(Circular Error Probable,CEP)的毁伤概率矩阵库(Damage Probability Matrix Library,DPML)和改进启发式退火优化机...针对形状不规则复杂面目标多弹瞄准点优化算法计算效率低、稳定性差、优化能力不足的问题,提出一种基于弹药圆概率偏差(Circular Error Probable,CEP)的毁伤概率矩阵库(Damage Probability Matrix Library,DPML)和改进启发式退火优化机制的高效瞄准点优化算法(Efficient Aiming Point Optimization Algorithm,EAPOA)。构建多弹瞄准点优化模型时,除考虑目标形状、导弹毁伤能力外,还考虑导弹直接毁伤、间接毁伤和多弹种联合毁伤等复杂因素对目标毁伤效果的影响。提出一种基于DPML的毁伤概率快速估计算法,提升算法优化效率和鲁棒性;设计一种基于候选瞄准点序列化的优化算法框架,并提出基于全局搜索和改进退火机制的启发式优化算法,降低瞄准点组合求解空间大小并提升算法优化能力。通过6个复杂面目标测试用例验证算法性能。研究结果表明,所提的EAPOA相比于增强精英保留策略遗传算法具有更强的优化能力,且平均优化时间仅为其1/5~1/3,在优化收益和计算效率上具有明显优势。展开更多
在电力巡检过程中,无人机等边端智能检测设备往往面临输电线路绝缘子缺陷目标小、背景因素复杂等难点,且边端设备的硬件条件限制了模型的规模,导致设备算力有限,模型准确率偏低。针对上述问题,该文提出了一种基于YOLOv8-RFL(You only lo...在电力巡检过程中,无人机等边端智能检测设备往往面临输电线路绝缘子缺陷目标小、背景因素复杂等难点,且边端设备的硬件条件限制了模型的规模,导致设备算力有限,模型准确率偏低。针对上述问题,该文提出了一种基于YOLOv8-RFL(You only look once version 8-RFL)模型的输电线路绝缘子缺陷检测方法。首先,通过对原有主干网络C2f(CSPDarknet53 to 2-Stage FPN)模块进行改进,增强模型对于绝缘子缺陷的特征提取能力;其次,构建基于特征聚焦的泛化特征金字塔网络(focusing generalized feature pyramid networks,FGFPN),采用“特征聚焦-扩散”的思想,精细化小缺陷目标的特征表达;然后,设计基于交叉注意机制的特征语义融合模块(feature semantic fusion module,FSFM),优化了对关键特征信息的捕获和利用;最后,提出轻量化权重共享检测头(Lightweight weight sharing detection head,LWSD),在保证检测精度的同时提高模型的计算效率和实时性。实验表明,改进后的YOLOv8-RFL模型均值平均精度(mean average precision,mAP)达到了93.2%,相较于基准模型提升了5.9%,在降低模型参数量和所需计算量的同时,实现了更好的绝缘子小目标缺陷检测效果,对于复杂背景下的输电线路绝缘子缺陷检测具有一定的现实意义。展开更多
文摘A simulation approach based on a full-wave numerical method is presented to study electromagnetic characteristics by complex targets. How to validate simulation results is considered thoroughly under no analytical and measured data, where a double-check criterion is designed for our simulation approach. As an example, the scattering of F-117A is studied by using our simulation approach under all polarizations, different frequency bands, incident and scattering directions, etc., some of which, such as cross-polarization, bistatic RCS, have not been considered in the previous literature.
文摘The electromagnetic scattering computation has developed rapidly for many years; some computing problems for complex and coated targets cannot be solved by using the existing theory and computing models. A computing model based on data is established for making up the insufficiency of theoretic models. Based on the ＂support vector regression method＂, which is formulated on the principle of minimizing a structural risk, a data model to predicate the unknown radar cross section of some appointed targets is given. Comparison between the actual data and the results of this predicting model based on support vector regression method proved that the support vector regression method is workable and with a comparative precision.
文摘A practical approach for calculating the RCS (Radar Cross Section) of complex targets modeled with wire-grid-frame is presented. A way for generating a polyhedron model (facet-wedge model) with the wire-grid-frame data is described. For storing and reading the data of the polyhedron model in an easy way, a data structure is given.
文摘针对形状不规则复杂面目标多弹瞄准点优化算法计算效率低、稳定性差、优化能力不足的问题,提出一种基于弹药圆概率偏差(Circular Error Probable,CEP)的毁伤概率矩阵库(Damage Probability Matrix Library,DPML)和改进启发式退火优化机制的高效瞄准点优化算法(Efficient Aiming Point Optimization Algorithm,EAPOA)。构建多弹瞄准点优化模型时,除考虑目标形状、导弹毁伤能力外,还考虑导弹直接毁伤、间接毁伤和多弹种联合毁伤等复杂因素对目标毁伤效果的影响。提出一种基于DPML的毁伤概率快速估计算法,提升算法优化效率和鲁棒性;设计一种基于候选瞄准点序列化的优化算法框架,并提出基于全局搜索和改进退火机制的启发式优化算法,降低瞄准点组合求解空间大小并提升算法优化能力。通过6个复杂面目标测试用例验证算法性能。研究结果表明,所提的EAPOA相比于增强精英保留策略遗传算法具有更强的优化能力,且平均优化时间仅为其1/5~1/3,在优化收益和计算效率上具有明显优势。
文摘在电力巡检过程中,无人机等边端智能检测设备往往面临输电线路绝缘子缺陷目标小、背景因素复杂等难点,且边端设备的硬件条件限制了模型的规模,导致设备算力有限,模型准确率偏低。针对上述问题,该文提出了一种基于YOLOv8-RFL(You only look once version 8-RFL)模型的输电线路绝缘子缺陷检测方法。首先,通过对原有主干网络C2f(CSPDarknet53 to 2-Stage FPN)模块进行改进,增强模型对于绝缘子缺陷的特征提取能力;其次,构建基于特征聚焦的泛化特征金字塔网络(focusing generalized feature pyramid networks,FGFPN),采用“特征聚焦-扩散”的思想,精细化小缺陷目标的特征表达;然后,设计基于交叉注意机制的特征语义融合模块(feature semantic fusion module,FSFM),优化了对关键特征信息的捕获和利用;最后,提出轻量化权重共享检测头(Lightweight weight sharing detection head,LWSD),在保证检测精度的同时提高模型的计算效率和实时性。实验表明,改进后的YOLOv8-RFL模型均值平均精度(mean average precision,mAP)达到了93.2%,相较于基准模型提升了5.9%,在降低模型参数量和所需计算量的同时,实现了更好的绝缘子小目标缺陷检测效果,对于复杂背景下的输电线路绝缘子缺陷检测具有一定的现实意义。
