基于改进的局部二值拟合主动轮廓模型的指静脉图像分割

对应用于图像分割的局部二值拟合(LBF)模型进行改进,提出一种加权全变分局部二值拟合能量泛函分割模型。改进后的模型在加权全变分分量中加入边缘停止函数,使得模型对边缘信息更加敏感,解决了指静脉图像对比度低、不易分割的问题。同时模型中采用正则化约束项,避免水平集函数的重新初始化,缩短了演化时间。对合成图像和真实指静脉图像的分割实验结果表明,本文模型比LBF模型具有更强的边缘细节分割能力,且不易陷入局部收敛,更加适用于指静脉图像的分割。     

Mid-infrared carbon monoxide detection system using differential absorption spectroscopy technique

A differential carbon monoxide (CO) concentration sensing device using a self-fabricated spherical mirror (e.g. light-collector) and a multi-pass gas-chamber is presented in this paper. Single-source dual-channel detection method is adopted to suppress the interferences from light source, optical path and environmental changes. Detection principle of the device is described, and both the optical part and the electrical part are developed. Experiments are carried out to evaluate the sensing performance on CO concentration. The results indicate that at 1.013×105 Pa and 298 K, the limit of detection (LoD) is about 11.5 mg/m3 with an absorption length of 40 cm. As the gas concentration gets larger than 115 mg/m3 (1.013×105 Pa, 298 K), the relative detection error falls into the range of ?1.7%—+1.9%. Based on 12 h long-term measurement on the 115 mg/m3 and 1 150 mg/m3 CO samples, the maximum detection errors are about 0.9% and 5.5%, respectively. Due to the low cost and competitive characteristics, the proposed device shows potential applications in CO detection in the circumstances of coal-mine production and environmental protection.     

基于百度飞桨AI Studio的机器学习教学新模式实践与探索

针对机器学习课程教学中存在的问题,提出以百度飞桨AI Studio平台为载体的线上线下混合式理论教学模式和基于项目式的实验教学模式,介绍具体教学改革方案与实践,为学生进一步学习人工智能其他课程奠定坚实的基础.     

H.264标准中Exp-Golomb编解码器的VLSI设计

为提高Exp-Golomb码的编解码效率,提出了一种基于快速“首位1检测”的Exp-Golomb编解码器硬件实现方法,降低了计算量并节省了硬件资源。该Exp-Golomb编解码器已通过RTL（Register Transfer Level）级仿真和综合,并在FPGA（Field Programmable Gate Array）开发平台进行了验证,在133 MHz时钟频率下编解码器的综合门数分别为765门和632门。该编解码器能满足Baseline档次（30帧/s）,分辨率为352×288视频序列的实时编解码对质量和速度的要求。     

仿真实验在“高频电子电路”课程中的作用

高频电子电路结构复杂,理论分析较难,实验受实际电路、频率较高等因素的影响,有些实验结果不易观察得到。针对这些问题,在理论教学和实验教学中采用不同的教学模式,有助于学生对理论知识的理解和运用,有利于学生对实验结果的分析和探讨。在教学过程中采用仿真实验辅助教学模式,能够激发学生的学习兴趣,提高学生自主学习的能力,进而提高教学效率,提升教学效果。     

一种可重构的H.264帧内预测硬件结构

H.264中的帧内预测是宏块解码的一部分,用来去除当前图像中的空间冗余,提高编码效率.本文根据帧内预测的特点,提出了一种合并运算的可重构硬件解码结构,将不同预测模式的相同运算进行整合,提高硬件模块的可重用性,在压缩硬件面积的同时,保证帧内预测的解码效率.基于上述方法实现的帧内预测模块已经通过了RTL级仿真和综合,在133MHz时钟频率下能够满足Baseline档次30帧/秒,分辨率为352x288视频序列的实时编解码对质量和速度的要求,被应用于符合H.264标准的网络视频会议平台中.     

适用于H264标准的快速去方块滤波结构

新一代视频压缩编码标准H.264引入环路滤波器来去除解码图像的方块效应,根据去方块滤波的特点,提出了一种环路滤波器的硬件实现结构,通过改进滤波算法,优化的缓冲区管理方式和流水线设计,提高了环路滤波器的处理效率和数据的吞吐能力.该硬件结构通过了RTL级仿真和综合,并在Spartan3 XC3S2000的FPGA平台上进行了验证,在133MHz工作频率下可以满足H.264标准Baseline档次30帧/秒分辨率为352×288标准视频序列的实时解码.     

A mid-infrared methane detection device based on dual- channel lock-in amplifier

A portable dual-channel digital/analogue hybrid lock-in amplifier (LIA) is developed, and its amplitude detection error is less than 10% when the signal-to-noise ratio (SNR) is larger than ?12 dB. Then, a differential mid-infrared methane (CH4) detection device is experimentally demonstrated based on a wideband incandescence wire-source and a multi-pass spherical reflector. The experiments are carried out to obtain the sensing performance of the device. With the absorption length of only ~4.8 cm, the limit of detection (LoD) is about 71.43 mg/m3, and the detection range is from 0 mg/m3 to 5.00×104 mg/m3. As the concentration gets larger than 714.30 mg/m3, the relative detection error falls into the range of ?5%—+5%. Two seven-hour-measurements are done on the CH4 samples with concentrations of 1.43×103 mg/m3 and 4.29×103 mg/m3, respectively, and the results show that the maximum relative error is less than 5%. Because of the cost effective incandescence wire-source, the small-size and inexpensive dual-channel LIA, and the small-size absorption pool and reflector, the developed device shows potential applications of CH4 detection in coal mine production and environmental protection.     

压缩感知的指静脉图像去噪

基于压缩感知理论的梯度投影稀疏重建(GPSR)算法对合成指静脉图像进行去噪预处理,运用Canny算子提取指静脉边缘验证了GPSR算法的去噪效果。实验结果表明,与全变分去噪算法(ROF去噪算法)相比,运用GPSR算法可以得到更高信噪比的指静脉图像、更清晰的指静脉边缘轮廓,解决了红外传感器提取指静脉信息时存在的静脉边界模糊、不易分割及提取边缘等问题。