针对复杂危险环境下传统吸烟行为检测方法成本较高、实时性较差、识别率低、容易漏检和小目标误判率高等问题,提出一种基于改进YOLOv9的吸烟行为检测算法YOLO-SSDA。YOLO-SSDA通过引入注意力机制模块SimAM,加强骨干网络对香烟特征的提...针对复杂危险环境下传统吸烟行为检测方法成本较高、实时性较差、识别率低、容易漏检和小目标误判率高等问题,提出一种基于改进YOLOv9的吸烟行为检测算法YOLO-SSDA。YOLO-SSDA通过引入注意力机制模块SimAM,加强骨干网络对香烟特征的提取能力;利用Soft-NMS算法避免在过滤重叠定位框时出现漏检可能;利用距离关联算法计算香烟与人体之间的距离关系,减少香烟识别误判率。本文通过网络采样、AI生成与现场采集等方式构建了实验数据集并进行了大量实验测试,实验结果表明:YOLO-SSDA算法准确率为94%,召回率为89.6%,mAP50值为94.4%,处理单振图像输入的平均耗时为39 ms,与其他算法相比具有较高的检测准确率和实时性能。Aiming at the problems of high cost, poor real-time performance, low recognition rate, easy missed detection, and high misjudgment rate of small targets in traditional smoking behavior detection methods in complex and dangerous environments, a smoking behavior detection algorithm YOLO-SSDA based on improved YOLOv9 is proposed. By introducing the attention mechanism module SimAM, the backbone network’s ability to extract cigarette features is strengthened;using Soft NMS algorithm to avoid the possibility of missed detection when filtering overlapping positioning boxes;using distance correlation algorithm to calculate the distance relationship between cigarettes and the human body, in order to reduce the misjudgment rate of cigarette recognition. This article constructed an experimental dataset through network sampling, AI generation, and on-site collection, and conducted extensive testing based on it. The results show that the YOLO-SSDA algorithm proposed in this paper has an accuracy of 94%, a recall rate of 89.6%, an mAP50 value of 94.4%, and an average processing time of 39 ms for single vibration image input. Compared with other algorithms, it has higher detection accuracy and real-time performance.展开更多
为实现SSD算法模型对人脸的目标检测,采用公开人脸数据集FDDB对网络模型进行重新训练改进。通过训练时输入不同的人脸数据集来优化网络训练结果。针对人脸检测训练过程中的过拟合问题,通过降噪自编码器的方法,在输入数据集中加入负样本...为实现SSD算法模型对人脸的目标检测,采用公开人脸数据集FDDB对网络模型进行重新训练改进。通过训练时输入不同的人脸数据集来优化网络训练结果。针对人脸检测训练过程中的过拟合问题,通过降噪自编码器的方法,在输入数据集中加入负样本,在训练模型中生成噪声。通过L1正则化产出稀疏模型,稀疏模型具有更好的特性去处理高维的数据特征以增强模型的泛化能力,实现在网络迭代训练过程中降噪的效果,防止模型陷入过拟合。然后通过非极大值抑制算法(NMS)使候选框确定为最终的人脸检测窗口进行人脸检测。在训练平台MXnet下的实验结果表明,加入噪声后的人脸检测模型的mAp(mean average precision)性能提高至0.997,同时在提高遮挡、光照、小目标等检测的鲁棒性的情况下,仍保持较快的收敛速度。展开更多
Considering that the hardware implementation of the normalized minimum sum(NMS)decoding algorithm for low-density parity-check(LDPC)code is difficult due to the uncertainty of scale factor,an NMS decoding algorithm wi...Considering that the hardware implementation of the normalized minimum sum(NMS)decoding algorithm for low-density parity-check(LDPC)code is difficult due to the uncertainty of scale factor,an NMS decoding algorithm with variable scale factor is proposed for the near-earth space LDPC codes(8177,7154)in the consultative committee for space data systems(CCSDS)standard.The shift characteristics of field programmable gate array(FPGA)is used to optimize the quantization data of check nodes,and finally the function of LDPC decoder is realized.The simulation and experimental results show that the designed FPGA-based LDPC decoder adopts the scaling factor in the NMS decoding algorithm to improve the decoding performance,simplify the hardware structure,accelerate the convergence speed and improve the error correction ability.展开更多
文摘针对复杂危险环境下传统吸烟行为检测方法成本较高、实时性较差、识别率低、容易漏检和小目标误判率高等问题,提出一种基于改进YOLOv9的吸烟行为检测算法YOLO-SSDA。YOLO-SSDA通过引入注意力机制模块SimAM,加强骨干网络对香烟特征的提取能力;利用Soft-NMS算法避免在过滤重叠定位框时出现漏检可能;利用距离关联算法计算香烟与人体之间的距离关系,减少香烟识别误判率。本文通过网络采样、AI生成与现场采集等方式构建了实验数据集并进行了大量实验测试,实验结果表明:YOLO-SSDA算法准确率为94%,召回率为89.6%,mAP50值为94.4%,处理单振图像输入的平均耗时为39 ms,与其他算法相比具有较高的检测准确率和实时性能。Aiming at the problems of high cost, poor real-time performance, low recognition rate, easy missed detection, and high misjudgment rate of small targets in traditional smoking behavior detection methods in complex and dangerous environments, a smoking behavior detection algorithm YOLO-SSDA based on improved YOLOv9 is proposed. By introducing the attention mechanism module SimAM, the backbone network’s ability to extract cigarette features is strengthened;using Soft NMS algorithm to avoid the possibility of missed detection when filtering overlapping positioning boxes;using distance correlation algorithm to calculate the distance relationship between cigarettes and the human body, in order to reduce the misjudgment rate of cigarette recognition. This article constructed an experimental dataset through network sampling, AI generation, and on-site collection, and conducted extensive testing based on it. The results show that the YOLO-SSDA algorithm proposed in this paper has an accuracy of 94%, a recall rate of 89.6%, an mAP50 value of 94.4%, and an average processing time of 39 ms for single vibration image input. Compared with other algorithms, it has higher detection accuracy and real-time performance.
文摘为实现SSD算法模型对人脸的目标检测,采用公开人脸数据集FDDB对网络模型进行重新训练改进。通过训练时输入不同的人脸数据集来优化网络训练结果。针对人脸检测训练过程中的过拟合问题,通过降噪自编码器的方法,在输入数据集中加入负样本,在训练模型中生成噪声。通过L1正则化产出稀疏模型,稀疏模型具有更好的特性去处理高维的数据特征以增强模型的泛化能力,实现在网络迭代训练过程中降噪的效果,防止模型陷入过拟合。然后通过非极大值抑制算法(NMS)使候选框确定为最终的人脸检测窗口进行人脸检测。在训练平台MXnet下的实验结果表明,加入噪声后的人脸检测模型的mAp(mean average precision)性能提高至0.997,同时在提高遮挡、光照、小目标等检测的鲁棒性的情况下,仍保持较快的收敛速度。
文摘Considering that the hardware implementation of the normalized minimum sum(NMS)decoding algorithm for low-density parity-check(LDPC)code is difficult due to the uncertainty of scale factor,an NMS decoding algorithm with variable scale factor is proposed for the near-earth space LDPC codes(8177,7154)in the consultative committee for space data systems(CCSDS)standard.The shift characteristics of field programmable gate array(FPGA)is used to optimize the quantization data of check nodes,and finally the function of LDPC decoder is realized.The simulation and experimental results show that the designed FPGA-based LDPC decoder adopts the scaling factor in the NMS decoding algorithm to improve the decoding performance,simplify the hardware structure,accelerate the convergence speed and improve the error correction ability.
