In order to find better simplicity measurements for 3D object recognition, a new set of local regularities is developed and tested in a stepwise 3D reconstruction method, including localized minimizing standard deviat...In order to find better simplicity measurements for 3D object recognition, a new set of local regularities is developed and tested in a stepwise 3D reconstruction method, including localized minimizing standard deviation of angles（L-MSDA）, localized minimizing standard deviation of segment magnitudes（L-MSDSM）, localized minimum standard deviation of areas of child faces （L-MSDAF）, localized minimum sum of segment magnitudes of common edges （L-MSSM）, and localized minimum sum of areas of child face （L-MSAF）. Based on their effectiveness measurements in terms of form and size distortions, it is found that when two local regularities： L-MSDA and L-MSDSM are combined together, they can produce better performance. In addition, the best weightings for them to work together are identified as 10% for L-MSDSM and 90% for L-MSDA. The test results show that the combined usage of L-MSDA and L-MSDSM with identified weightings has a potential to be applied in other optimization based 3D recognition methods to improve their efficacy and robustness.展开更多
3D objects can be stored in computer of different describing ways, such as point set, polyline, polygonal surface and Euclidean distance map. Moment invariants of different orders may have the different magnitude. A m...3D objects can be stored in computer of different describing ways, such as point set, polyline, polygonal surface and Euclidean distance map. Moment invariants of different orders may have the different magnitude. A method for normalizing moments of 3D objects is proposed, which can set the values of moments of different orders roughly in the same range and be applied to different 3D data formats universally. Then accurate computation of moments for several objects is presented and experiments show that this kind of normalization is very useful for moment invariants in 3D objects analysis and recognition.展开更多
为实现YOLO(you only look once)模型的超参数自动优化,提出基于正交优化策略的YOLO模型超参数优化方法(hyper-parameter optimization of YOLO model based on orthogonal optimization strategy,OOS)。首先基于统计学的正交试验原理,...为实现YOLO(you only look once)模型的超参数自动优化,提出基于正交优化策略的YOLO模型超参数优化方法(hyper-parameter optimization of YOLO model based on orthogonal optimization strategy,OOS)。首先基于统计学的正交试验原理,提出了种群的正交搜索方法与超参数贡献度分析策略,提高了算法的优化效率;然后,设计了均匀正交搜索策略和邻域正交搜索策略,以缓解YOLO模型陷入局部最优和早熟收敛问题。最后,在NWPU VHR-10和Pascal VOC两个目标检测数据集上,以YOLOv5、YOLOv5s-Transformer和YOLOv7为优化对象进行测试,测试结果表明,所提出的OOS超参数优化方法对于YOLO模型的识别精度均有所提升。在两个数据集上的平均识别精度mAP@0.5分别提升至93.94%、93.18%、93.45%以及85.81%、84.59%、90.62%;mAP@0.5-0.95提升至60.00%、60.08%、56.98%以及62.27%、58.89%、71.91%,可为目标检测模型的超参数智能优化提供一种新方法。展开更多
文摘In order to find better simplicity measurements for 3D object recognition, a new set of local regularities is developed and tested in a stepwise 3D reconstruction method, including localized minimizing standard deviation of angles（L-MSDA）, localized minimizing standard deviation of segment magnitudes（L-MSDSM）, localized minimum standard deviation of areas of child faces （L-MSDAF）, localized minimum sum of segment magnitudes of common edges （L-MSSM）, and localized minimum sum of areas of child face （L-MSAF）. Based on their effectiveness measurements in terms of form and size distortions, it is found that when two local regularities： L-MSDA and L-MSDSM are combined together, they can produce better performance. In addition, the best weightings for them to work together are identified as 10% for L-MSDSM and 90% for L-MSDA. The test results show that the combined usage of L-MSDA and L-MSDSM with identified weightings has a potential to be applied in other optimization based 3D recognition methods to improve their efficacy and robustness.
基金Supported by National Key Basic Research Program(No.2004CB318006)National Natural Science Foundation of China(Nos.60873164,60573154,60533090,61379082 and 61227802)
文摘3D objects can be stored in computer of different describing ways, such as point set, polyline, polygonal surface and Euclidean distance map. Moment invariants of different orders may have the different magnitude. A method for normalizing moments of 3D objects is proposed, which can set the values of moments of different orders roughly in the same range and be applied to different 3D data formats universally. Then accurate computation of moments for several objects is presented and experiments show that this kind of normalization is very useful for moment invariants in 3D objects analysis and recognition.
文摘为实现YOLO(you only look once)模型的超参数自动优化,提出基于正交优化策略的YOLO模型超参数优化方法(hyper-parameter optimization of YOLO model based on orthogonal optimization strategy,OOS)。首先基于统计学的正交试验原理,提出了种群的正交搜索方法与超参数贡献度分析策略,提高了算法的优化效率;然后,设计了均匀正交搜索策略和邻域正交搜索策略,以缓解YOLO模型陷入局部最优和早熟收敛问题。最后,在NWPU VHR-10和Pascal VOC两个目标检测数据集上,以YOLOv5、YOLOv5s-Transformer和YOLOv7为优化对象进行测试,测试结果表明,所提出的OOS超参数优化方法对于YOLO模型的识别精度均有所提升。在两个数据集上的平均识别精度mAP@0.5分别提升至93.94%、93.18%、93.45%以及85.81%、84.59%、90.62%;mAP@0.5-0.95提升至60.00%、60.08%、56.98%以及62.27%、58.89%、71.91%,可为目标检测模型的超参数智能优化提供一种新方法。
