Sonar image processing system is an important intelligent system of Autonomous Un-derwater Vehicle.Based on TMS320C30 high speed DSP,it is used to realize sonar imagecompression and underwater object detections includ...Sonar image processing system is an important intelligent system of Autonomous Un-derwater Vehicle.Based on TMS320C30 high speed DSP,it is used to realize sonar imagecompression and underwater object detections including obstacle recognition in real time.Inthis paper,the software and hardware designs of this system are introduced and the experi-mental results are given.展开更多
In order to achieve high-speed, real-time and accurate, an image acquisition method based on digital signal processor (DSP) TMS320DM642 is proposed for the paper currency image acquisition [1]. System will be high spe...In order to achieve high-speed, real-time and accurate, an image acquisition method based on digital signal processor (DSP) TMS320DM642 is proposed for the paper currency image acquisition [1]. System will be high speed digital signal processing (DSP) technology and complex programmable logic device (CPLD) and CIS acquisition module combination, the structure of acquisition system is given and the time series analysis, during the process of collecting this kind of design has the advantages of simple implementation, high recognition rate [2].展开更多
By utilizing the capability of high-speed computing,powerful real-time processing of TMS320F2812 DSP,wavelet thresholding denoising algorithm is realized based on Digital Signal Processors.Based on the multi-resolutio...By utilizing the capability of high-speed computing,powerful real-time processing of TMS320F2812 DSP,wavelet thresholding denoising algorithm is realized based on Digital Signal Processors.Based on the multi-resolution analysis of wavelet transformation,this paper proposes a new thresholding function,to some extent,to overcome the shortcomings of discontinuity in hard-thresholding function and bias in soft-thresholding function.The threshold value can be abtained adaptively according to the characteristics of wavelet coefficients of each layer by adopting adaptive threshold algorithm and then the noise is removed.The simulation results show that the improved thresholding function and the adaptive threshold algorithm have a good effect on denoising and meet the criteria of smoothness and similarity between the original signal and denoising signal.展开更多
In this paper, we present an optimized design method for high-speed embedded image processing system using 32 bit floating-point Digital Signal Processor （DSP） and Complex Programmable Logic Device （CPLD）. The DSP...In this paper, we present an optimized design method for high-speed embedded image processing system using 32 bit floating-point Digital Signal Processor （DSP） and Complex Programmable Logic Device （CPLD）. The DSP acts as the main processor of the system： executes digital image processing algorithms and operates other devices such as image sensor and CPLD. The CPLD is used to acquire images and achieve complex logic control of the whole system. Some key technologies are introduced to enhance the performance of our system. In particular, the use of DSP/BIOS tool to develop DSP applications makes our program run much more efficiently. As a result, this system can provide an excellent computing platform not only for executing complex image processing algorithms, but also for other digital signal processing or multi-channel data collection by choosing different sensors or Analog-to-Digital （A/D） converters.展开更多
矩阵转置是矩阵运算的基本操作,广泛应用于信号处理、科学计算以及深度学习等各种领域。随着国防科技大学自主研制的飞腾异构多核数字信号处理器(digital signal processor, DSP)在各种领域中的推广应用,对高性能矩阵转置实现提出了强...矩阵转置是矩阵运算的基本操作,广泛应用于信号处理、科学计算以及深度学习等各种领域。随着国防科技大学自主研制的飞腾异构多核数字信号处理器(digital signal processor, DSP)在各种领域中的推广应用,对高性能矩阵转置实现提出了强烈需求。针对飞腾异构多核DSP的体系结构特征与矩阵转置操作的特点,提出了一种适配不同数据位宽(8 B、4 B以及2 B)矩阵的并行矩阵转置算法ftmMT。该算法基于DSP中向量处理单元的Load/Store部件实现了向量化,同时基于矩阵分块实现了多个DSP核的并行处理,通过隐式乒乓设计实现了片上向量化转置与片外访存的重叠以及访存性能的大幅提升。实验结果表明,ftmMT能够显著加快矩阵转置操作,与CPU上的开源转置库HPTT相比,可获得高达8.99倍的性能加速。展开更多
矩阵乘卷积算法能够为各种卷积配置提供高性能基础实现,是面向给定芯片进行卷积性能优化的首要选择。针对国防科技大学自主研制的飞腾异构多核数字信号处理器(digital signal processor,DSP)芯片的特征以及矩阵乘卷积算法自身的特点,提...矩阵乘卷积算法能够为各种卷积配置提供高性能基础实现,是面向给定芯片进行卷积性能优化的首要选择。针对国防科技大学自主研制的飞腾异构多核数字信号处理器(digital signal processor,DSP)芯片的特征以及矩阵乘卷积算法自身的特点,提出了一种面向多核DSP架构的高性能并行矩阵乘卷积实现算法ftmEConv。该算法由输入特征图转换、卷积核转换、矩阵乘以及输出特征图转换这四个均运行在通用多核DSP上的并行化部分构成,通过有效挖掘通用DSP核中功能单元的潜力来提升各个部分的性能。实验结果表明,ftmEConv实现了高达42.90%的计算效率,与芯片上的其他矩阵乘卷积算法实现相比,获得了高达7.79倍的性能加速。展开更多
数字图像处理是计算机视觉和计算机图形学的重要分支,已经成为当下广泛应用的研究领域。研究旨在探索数字图像处理技术及其在计算机视觉领域中的应用,将深入研究数字图像的增强、分割、识别与分类等处理方法,为实现图像自动化分析提供...数字图像处理是计算机视觉和计算机图形学的重要分支,已经成为当下广泛应用的研究领域。研究旨在探索数字图像处理技术及其在计算机视觉领域中的应用,将深入研究数字图像的增强、分割、识别与分类等处理方法,为实现图像自动化分析提供技术支撑。研究表明,在数字图像增强处理中,采用直方图均衡化、灰度化等方法可以明显提高图片质量,在分割处理中,采用聚类算法和分水岭算法可以较好地实现图像的分割,识别与分类方面,采用尺度不变特征转换(scale-invariant feature transform,SIFT)、方向性FAST特征点检测和旋转BRIEF描述子(oriented FAST and rotated BRIEF,ORB)算法等可以更好地对图片进行识别和分类。展开更多
基金the High Technology Research and Development Programme of china.
文摘Sonar image processing system is an important intelligent system of Autonomous Un-derwater Vehicle.Based on TMS320C30 high speed DSP,it is used to realize sonar imagecompression and underwater object detections including obstacle recognition in real time.Inthis paper,the software and hardware designs of this system are introduced and the experi-mental results are given.
文摘In order to achieve high-speed, real-time and accurate, an image acquisition method based on digital signal processor (DSP) TMS320DM642 is proposed for the paper currency image acquisition [1]. System will be high speed digital signal processing (DSP) technology and complex programmable logic device (CPLD) and CIS acquisition module combination, the structure of acquisition system is given and the time series analysis, during the process of collecting this kind of design has the advantages of simple implementation, high recognition rate [2].
文摘By utilizing the capability of high-speed computing,powerful real-time processing of TMS320F2812 DSP,wavelet thresholding denoising algorithm is realized based on Digital Signal Processors.Based on the multi-resolution analysis of wavelet transformation,this paper proposes a new thresholding function,to some extent,to overcome the shortcomings of discontinuity in hard-thresholding function and bias in soft-thresholding function.The threshold value can be abtained adaptively according to the characteristics of wavelet coefficients of each layer by adopting adaptive threshold algorithm and then the noise is removed.The simulation results show that the improved thresholding function and the adaptive threshold algorithm have a good effect on denoising and meet the criteria of smoothness and similarity between the original signal and denoising signal.
基金Supported by the National Natural Science Foundation of China (No.60472046)
文摘In this paper, we present an optimized design method for high-speed embedded image processing system using 32 bit floating-point Digital Signal Processor （DSP） and Complex Programmable Logic Device （CPLD）. The DSP acts as the main processor of the system： executes digital image processing algorithms and operates other devices such as image sensor and CPLD. The CPLD is used to acquire images and achieve complex logic control of the whole system. Some key technologies are introduced to enhance the performance of our system. In particular, the use of DSP/BIOS tool to develop DSP applications makes our program run much more efficiently. As a result, this system can provide an excellent computing platform not only for executing complex image processing algorithms, but also for other digital signal processing or multi-channel data collection by choosing different sensors or Analog-to-Digital （A/D） converters.
文摘矩阵乘卷积算法能够为各种卷积配置提供高性能基础实现,是面向给定芯片进行卷积性能优化的首要选择。针对国防科技大学自主研制的飞腾异构多核数字信号处理器(digital signal processor,DSP)芯片的特征以及矩阵乘卷积算法自身的特点,提出了一种面向多核DSP架构的高性能并行矩阵乘卷积实现算法ftmEConv。该算法由输入特征图转换、卷积核转换、矩阵乘以及输出特征图转换这四个均运行在通用多核DSP上的并行化部分构成,通过有效挖掘通用DSP核中功能单元的潜力来提升各个部分的性能。实验结果表明,ftmEConv实现了高达42.90%的计算效率,与芯片上的其他矩阵乘卷积算法实现相比,获得了高达7.79倍的性能加速。
文摘数字图像处理是计算机视觉和计算机图形学的重要分支,已经成为当下广泛应用的研究领域。研究旨在探索数字图像处理技术及其在计算机视觉领域中的应用,将深入研究数字图像的增强、分割、识别与分类等处理方法,为实现图像自动化分析提供技术支撑。研究表明,在数字图像增强处理中,采用直方图均衡化、灰度化等方法可以明显提高图片质量,在分割处理中,采用聚类算法和分水岭算法可以较好地实现图像的分割,识别与分类方面,采用尺度不变特征转换(scale-invariant feature transform,SIFT)、方向性FAST特征点检测和旋转BRIEF描述子(oriented FAST and rotated BRIEF,ORB)算法等可以更好地对图片进行识别和分类。
