双目视觉技术是重要的立体成像技术之一,该技术的核心是计算匹配像素点之间的视差值,而这种计算依赖于代价函数的设计.其中Absolute Difference Census(AD-Census)函数作为代价函数是常用的方法,但该方法受权重系数的影响较大.为了有效提高AD-Census函数的匹配精度,提出向AD-Census函数中添加截断参数以改进其匹配效果,保证了代价聚合时像素的绝对值和梯度不会出现因其中一个值较大,而导致另一个值的影响被忽略的情况出现,从而提高匹配精度,与传统的AD-Census函数相比失配率可以降低6％以上.

介绍"金课"的发展背景,探讨"金课"的内涵,剖析计算机通信与网络课程现状,并根据"金课"的内涵和建设要求,针对课程现状和存在的问题,从课程建设和课程思政两个角度,进行了"金课"建设的探索和思考.一方面,从课程目标、课程讲授内容、课程实施过程和课程评价4个角度,深入讨论了课程建设思路.另一方面,结合课程内容,从课程思政建设角度也进行了探索.

针对闪存存储单元会出现多位错误,提出了NAND型闪存存储器的二维汉明码算法,从行和列二维结构上对任意2J bit字长的存储数据,用统一的公式化算法进行汉明编码,通过计算纠错偶对能够纠错多位数据误码的情况,译码时先纠错,再校验,保证存储系统数据的完整性,解决了MLC结构闪存的纠错问题.在此基础上分析了不能纠错的图样模式.该算法核心结构上只有异或操作,易于实现,适用于NAND型闪存存储器.实验表明,闪存系统中有效数据二维结构为1 024×128时,采用二维汉明码结构,编码效率为11％,实际编码效率为12.6％.其编码速度和译码速度均达到了102 Gbps,适用于高速存储系统.

空间存储器存在高能粒子反转问题,提出了采用变长匹配的RS码纠错的大容量NAND型闪存图像存储器.为保证存储器实时纠错编译码时不影响持续读写速度,提出了主副总线的闪存和SRAM阵列备份的体系结构.在此基础上,针对图像数据源分辨率变化和NAND闪存是页存储体系的特点,提出了变长匹配的RS码字页编程方法,可纠错有限字节的随机错误,实现图像数据源、纠错码和闪存存储的匹配.针对突发长码错误,提出了固定周期的数据交织方法.固化参数后实验结果表明,在双总线系统架构下,能够保证系统存取速度,图像存储速率达到1.44 Gbps,读取速率达到1.2 Gbps,单页纠错能力突破30 byte,同误比特率下信噪比提高了2 dB以上.

There exist some problems in monitoring the flowrate of sewerage in drainage pipe, which include the difficult installation of devices and low accuracy of monitoring.To solve these problems,it was supposed to use the immerged double-beam doppler flowmeter for surveillance of sewerage flowrate.The installation method of this device could let it be installed easily in the drain-age pipe through inspection well.In addition,its double-beam doppler signal could remove the in-fluence of the vortex and turbulence in the pipe on the surveillance accuracy.This paper introduced the component,working principle,working process and installation requirements of the immerged double-beam doppler flowmeter as well as its practical engineering application.

There is no efficient way other than manual detection regarding maintenance and management of urban sewage pipe networks.To address this problem this paper proposes a system to detect flow of sewage pipe networks in cities based on plug-in multifactor probes.The system is able to improve the management of urban sewage discharges and lower the maintenance costs of sewage networks by means of automatic real-time detection.In application it can detect and warn any flow of sewage and spot fault points.

Any length of 2Jbit Boolean data to have the correct code which has the uniform structure and formula is provided .The algorithm is derived from the concept of the linear group code ,based on the deducing of generator matrix and the corresponding parity check matrix ,we conclude the algorithm can correct one bit data and identify two bit data .Then based on the analysis of the position of the data in the formula ,we get the information of the position of the error data . And we get the uniform formula . The experiment validates the correctness of the algorithm .The algorithm is adaptive to the high speed data storage and transfer .

The accuracy and efficiency as the two main evaluation indexes for k-means algorithm are influenced by the choice of initial clustering centers and the partition method of data points. In this paper, in view of the deficiency of direct k-means algorithm which chooses initial centers randomly, we propose a novel method about initial clustering centers based on sorting and partition and apply it to real data as well as simulated data, which show that this is an efficient method to improve the clustering accuracy and efficiency.

In k‐means clustering ,we are given a set of n data points in d‐dimensional space Rd and an integer K the problem is to determine a set of Kpoints in Rd ,called centers ,so as to minimize the mean squared distance from each data point to its nearest center .The initial centers of direct k‐means algorithm are chosen randomly ,different initial centers will lead to different results .In this paper ,in view of the deficiency of direct k‐means algorithm ,we propose a novel method about initial centers based on sorting and partition and apply it to real data as well as simulated data , which show that this is a simple and efficient method to improve the clustering accuracy and efficiency .

This paper puts forward some improved ideas on distributed video encoding scheme based on wavelet transform.First,a simplified algorithm is introduced to enhance the efficiency of the encoding side.A direct filling method for the high-frequency side information is then proposed to save the channel bandwidth.In the end,the idea of hierarchical transmission in traditional encoding is used to adapt to different transmission channel bandwidths.The simulation results show that this scheme can reduce transmission bit rate and guarantee high quality of video transmission.

As the important ground support equipment for aerial remote sensing instrument,the maintenance vehicle is used to overhaul and maintain aerial remote sensing instrument.In this paper,aimed at its requirement,the function of the maintenance vehicle is analyzed,and the maintenance vehicle with aided assembling and adjusting function is designed.A calculation method of static balance for the maintenance vehicle is proposed,and the theoretical thickness of balance mass is derived.Furthermore,the structure of the maintenance vehicle is analyzed with ANSYS,and the results show that the strength and stiffness of the maintenance vehicle meet the requirement.

This paper studies the principles of a 3D display system and shutter glasses control principle.The method of wireless radio frequency synchronization is designed with hardware block diagram and software flow pattern by using the method of wireless radio frequency synchronization.The system which presents a concrete method of calculating time delay compensation can solve the problems of time synchronization and frame synchronization because nRF24L01 wireless radio frequency chip is adopted to transmit synchronizing information.Finally,the practical effect of the glasses is tested and the graph of the effect is provided.

In this paper we use the whole frame loss recovery algorithm and propose a new method of improving the side information that improving the algorithm in order to adapt the nonlinear motion of image and estimate the side information in Distributed Video Coding.The result shows that the scheme can widely advance the accurate of the side information,reduce the bit-rate of the WZ frames and improve the performance of the Distributed Coding.

Disparity of stereo image pairs forms the basis of stereoscopic display, and the basic principle of binocular stel~eoscopic display is that the left and right images in stereo pairs with disparity are sent to the left eye and right eye respectively, resulting in a stereo sense. This paper presents a method to compute the disparity of stereo pairs in a vil~ual scene using ray tracing, by which disparity maps and depth maps are created. Besides, the characteristics of and factors affecting the disparity are analyzed.

The correlation between sequences is not made full use of in the fixed image group structure in distributed video coding system.Considering such an issue,this paper proposes an interpolation algorithm in which image group structure is controlled through self-adaptive adjustment at the coding end and self-adaptive image group at the decoding end.The algorithm estimates the correlation by matching inter-frame macro blocks.Dynamic frames are divided into key frame and Wyner-Ziv frame,aimed at fully using correlation between frames and improving the performance of the system.The experimental results show that the peak signal-to-noise ratio(PSNR) of side information increases by 0.1 to 1.3 dB.

According to the high and low frequency components of video sequence which contained different important information,the components of different frequencies will be encoded with different methods.The high and low frequency components of the frequency domain(Wyner-Ziv;WZ) frame be extracted and then doing non-equilibrium Codec based on the block respectively.For the distributed video coding schemes based on low-density parity-check accumulate codes(Low-Density Parity-Cheek Accumulate Codes;LDPCA),a non-equilibrium Codec method of he frequency domain frame was proposed.Compared with the traditional Codec scheme which the high and low frequency components are unified,the non-equilibrium Codec scheme will have better decoding performance.

The technology used to converse 2D to 3D can solve the problem of insufficiency of 3D stereoscopic videos because it is able to provide a huge number of videos of this kind by conversing 2D videos to 3D videos.As to such conversion,the key is to extract depth maps of high quality.In this connection,this paper puts forward a method for depth map extraction by integrating the characteristics of motion and texture.To estimate depth maps,foreground and background are firstly separated,and then foreground depth map is extracted based on motion,and background depth map based on texture characteristics.The final depth map is obtained by combining the two maps effectively.The results of the experiments show that depth maps that meet demands can be obtained by using the method proposed in this paper.

Fast 2-D Walsh Transformation algorithm is presented based on NVIDIA's GPU which support Compute Unified Device Architecture(CUDA).On the basis of the parallel architectureand hardware characteristic of GPU,the paper introduces three methods to improve the implementation performance:optimizing algorithm,texture Storage technology,and setting up logic Device Architecture.The experiment result shows that with the increasing of picture resolution,the runtime of 2-D Walsh Transformation based on GPU is far fewer than on the CPU.

Ray tracing based on acceleration structure includes two procedures: building and traversal the acceleration structure,kd-trees are considered by many to be the best known method to accelerate ray tracing of static scenes.The acceleration structures are crucial for fast ray tracing and need to be rebuilt or updated as the scene changes. Therefore,the structure rebuilding becomes a bottleneck during ray tracing procedure. A novel algorithm is proposed to build kd-trees for static and moving objects respectively, and it only needs to rebuild or update the dynamic kd-tree while keep the static kd- tree stable.We implement this fast rendering for dynamic scene at 3 ～5 fps for 1 024 × 1 024 images on CPU+GPU platform.

KD-tree is a traditional accelerating algorithm in 3D scene rendering,and its parallel processing performance is restricted to the unsupported recursion operations on SIMD architecture.A novel parallel algorithm is proposed to speed up KD-tree traversal on GPU.Clues pointing to six-direction subsequent nodes are inserted into the KD-tree structure.Thus,the algorithm efficiency is improved by avoiding stack data structure and reducing the worthless operations.The experiment results verify that the clued KD-tree algorithm increases the rendering speed by 3 to 8 times than the traditional algorithms.