Search

End-to-end person search is a novel task that integrates pedestrian detection and person re-identification (re-ID) into a joint optimization framework. However, the pedestrian features learned by most existing methods are not discriminative enough due to the potential adverse interaction between detection and re-ID tasks and the lack of discriminative power of re-ID loss. To this end, we propose an Improved Model Structure (IMS) with a novel re-ID loss function called Cosine Margin Online Instance Matching (CM-OIM) loss. Firstly, we design a model structure more suitable for person search, which alleviates the adverse interaction between the detection and re-ID parts by reasonably decreasing the network layers shared by them. Then, we conduct a full investigation of the weight of re-ID loss, which we argue plays an important role in end-to-end person search models. Finally, we improve the Online Instance Matching (OIM) loss by adopting a more robust online update strategy, and importing a cosine margin into it to increase the intra-class compactness of the features learned. Extensive experiments on two challenging datasets CUHK-SYSU and PRW demonstrate our approach outperforms the state-of-the-arts.

Based on the ZYNQ platform, this paper proposes a hardware/software co-design approach, and implements a binocular stereo vision system with high real-time performance and good human-computer interaction, which can be used to assist advanced driver assistance systems to improve driving safety. Combining the application characteristics of binocular stereo vision, the approach firstly modularizes the system’s functions to perform hardware/software partitioning, accelerates the data processing on FPGA, and performs the data control on ARM cores; then uses the ARM instruction set to configure the registers within FPGA to design relevant interfaces to complete the data interaction between hardware and software; finally, combines the implementation of specific algorithms and logical control to complete the binocular stereo vision system. The test results show that the frame rate with an image resolution of 640 * 480 can reach 121.43 frames per second when the FPGA frequency is 100M, and the frame rate is also high for large resolution images. At the same time, the system can achieve real-time display and human-computer interaction with the control of the graphical user interface.

The fingerprint based indoor localization is becoming a dominant solution for its high applicability in complex indoor environment. However, the extensive site survey efforts on manpower and time have become a major bottleneck. Based on the crowdsourcing method, the paper puts forward a novel indoor localization with the fusion of RFID (radio frequency identification devices), PDR (pedestrian dead reckoning)and MM (magnetic matching)technologies. First, a zero-effort fingerprint automated construction and site survey update scheme is proposed with the dual-frequency RFID. Second, in order to solve the problem that step length would vary from person to person which results into positioning bias, the RSS technology and floor map is introduced to aid step length estimation. Third, the particle filter is adopted to fuse the advantages of three different technologies to conduct the accurate indoor positioning. In the experiment, the obtained fingerprint database is demonstrated to possess a comparable accuracy with the human-annotated database. Also, the experiment results show that the proposed method achieves the comparable positioning accuracy and the positioning accuracy is 40% higher than the PDR system or RFID positioning system alone.

With the continued requirement of internet of things (IoT) development, many researchers focused on developing access and routing algorithms optimization of wireless sensor networks (WSNs) for communications quality guarantee, but less research has focused on the optimization of WSNs based on the mobile cellular networks (MCNs) assist. In this paper, we propose a convergence architecture of WSN and MCN for IoT applications, especially on the smart user equipment (UE) acting as a mobile gateway for WSN data transmission. And then, we investigate system application requirement for IoT. Finally, we discuss the key technical challenges and research points based on an analysis of smart gateway.

A text map based indoor localization is proposed with the fusion of radio frequency identification (RFID) and inertial measurement unit (IMU) in the narrow corridor space. The floor plan in narrow corridor is abstracted to a text map which uses the form of characters to represent indoor physical map. Then the indoor localization would be changed in the process of text processing and matching. When tag carrier is walking in the positioning area, the character string would be constructed with pedestrian dead reckoning (PDR) technology according to the path information. We search the character string in the text map and get the corresponding candidate locations, and then received signal strength (RSS) based fingerprint matching method is used to obtain the accurate location from candidate locations. The experiment is conducted to show that proposed method can reduce the mean positioning error to around 1.2 m without initial location specification.

Data aggregation is an essential operation in many applications of Wireless Sensor Networks (WSNs). For energy constrained and delay sensitive WSNs, how to prolong the lifetime and reduce the latency of the network are two critical problems. We deal with these two problems together in data aggregation by proposing a Load Balanced and Delay Efficient Algorithm (LBDEA), which consists of two phases. First, construct a load-balanced data aggregation tree based on Semi-Matching algorithm in the graph theory. Second, seek a collision-free scheduling based on greedy strategy for the transmission links in the aggregation tree. Through balancing the load of the sensor nodes, the lifetime of the network is optimized. By avoiding collisions and retransmissions of data delivery, the total latency is decreased. Simulations show that LBDEA not only improves the lifetime but also cuts down the delay for data aggregation in WSNs, compared with other related algorithms.

By analyzing the law of the usage of historical spectrum data, spectrum prediction can be used to predict the future usage state of the spectrum. This technology can be taken to predict the jamming signal in communication countermeasure. In order to achieve the reliable forecast of the jamming signal in the artificial interference channel environment (a channel with the man-made interference), this paper proposes a spectrum prediction algorithm based on suffix array, which could guide the cognitive users to switch between spectrums autonomously. Moreover, a cognitive radio system is built to verify the algorithm. Experiments demonstrate that the proposed algorithm predicts well with the small sampled historical spectrum data.

This paper presents a simple single-carrier (SC) distributed space-time block coded transmission scheme for asynchronous cooperative communications over frequency-selective channels. The cyclic prefix (CP) at the source node and relay nodes is used to combat inter-symbol interference from multi-path fading and the timing errors from the relay nodes. In this scheme, the relay nodes only implement very simple operations on their received signals to construct distributed space-time block code, and decoding, DFT and IDFT operations are only needed at the destination node. The low complexity SC zero-forcing frequency domain equalization (ZF-FDE) is applied at the destination node to recover the transmitted signals. The analysis and simulation results demonstrate that this simple SC scheme can achieve order-two spatial diversity gain for amplify-and-forward based cooperative communication over frequency-selective channels in the presence of asynchronization between relay nodes.