Infrared Real—time Thermal System Based on DSP

An infrared real-time imaging system using DSP(digital signal processor)as the kernel of digital signal processing board is presented.In this system,the imaging difference and nonuniformity correction method is developed on the chip taking advantage of DSP with high speed.The method combines hardware and software together,so that the difficulty for realizing such a method with other hardware can be overcome.     

基于数学形态学和投影法的车牌定位方法

在车牌定位系统中,针对图像中存在的复杂背景、车辆自身的车灯或标志这些因素的影响,提出了车牌初步定位和精确定位的方法.首先,对车牌图像进行灰度化、灰度拉伸、边缘检测后,运用数学形态学运算和车牌自身的宽高比例初步定位出车牌区域.然后,运用投影法分别进行水平方向和垂直方向的定位,得到准确的车牌区域.该方法能够很好地排除干扰因素,精确定位出车牌.     

High Sensitivity FBG Pressure Sensor Based on Mechanical Amplifier and Interrogation Using Tunable FBG Filter

A high sensitivity fiber Bragg grating pressure sensor by using mechanical amplifier is demonstrated. The measured pressure sensitivity is -1.80×10 -4 /MPa, which is about two orders of magnitude better than a simple monomode fiber with an in-fiber grating. The resolution of pressure measurement is 0.015 MPa based on interrogation using tunable fiber grating filter.     

Formation of the Moving Analogy Target with C＋＋ Language in TV Tracking Equipment

Moving analogy target is a key component of the performance testing system in TV tracking equipment.A new method is provided to produce the moving analogy target whose motion speed,track,contrast and size can be varied.The video signal transformed by video switching card is used to test the performances of the electronic box of TV tracking equipment.These performances include minimal tracking contrast,minimal size of tracking target,maximal tracking speed and capture time.     

一种基于视觉感知的舰船目标智能化识别方法

为有效识别视觉系统采集的可见光图像中的舰船目标,提出了基于YOLO（You Only Look Once）网络模型改进的10层的卷积神经网络（Convolutional Neural Network,CNN）用于水面舰船目标的智能识别,通过反卷积的方法可视化CNN中不同卷积层提取到的舰船目标特征。按照传统目标识别方法提取了舰船目标的四类典型人工设计特征,将所提CNN的舰船目标识别结果与YOLO网络模型及四类人工设计特征结合支持向量机用于舰船目标识别的结果进行比较。实验结果表明,与YOLO网络模型相比,综合精确率、召回率和效率3个舰船目标识别的性能指标,改进后的CNN性能更好,从而验证了所提方法的有效性。不同数据量下采用典型特征识别舰船目标与基于深度CNN识别舰船目标的识别结果比较说明了不同类型目标识别算法的优劣势,有利于推动综合性视觉感知框架的构建。