Haze reduces visibility in outdoor images by lowering contrast and obscuring scene details. This degradation undermines the reliability of practical vision systems. Prior-based dehazing methods often struggle in highlight-dominant scenes because atmospheric-light estimation can be corrupted by non-sky bright regions. To address this issue, we propose a highlight-aware image dehazing framework based on sky-feature-constrained atmospheric-light estimation and its real-time ZYNQ implementation. The method identifies reliable sky cues from the dark channel using brightness consistency, texture flatness, color uniformity, and spatial location, and then performs adaptive transmission estimation, radiance recovery, and gamma correction within the dark-channel-prior framework. A PS-PL co-design on ZYNQ7020 is further developed for real-time video dehazing. On the SOTS-Outdoor benchmark, the proposed method achieves 19.68 dB PSNR and 0.8196 SSIM, outperforming the classical DCP baseline by 3.62 dB in PSNR, while the deployed system reaches 108 FPS on ZYNQ7020. These results indicate that the proposed framework provides a practical solution for scene-adaptive and deployable dehazing in adverse outdoor environments.

Real-time detection of marine organisms plays a critical role in underwater ecological monitoring, endangered species protection, and autonomous underwater vehicle (AUV) operations. However, the degraded underwater images with low contrast and detail blur and limited embedded computing resources make it challenging for existing methods to balance accuracy and real-time performance. To address these problems, this paper proposes LMOD-YOLO, a lightweight real-time marine organism detection model. The whole model integrates three key modules to improve computational efficiency while maintaining high detection accuracy. Cross Stage Local Partial Convolution (CSLPC) module replaces the original C2f module to reduce redundant computation, improving spatial feature extraction efficiency. An Adaptive Spatial-Channel Dual Convolution enhanced Cross-scale Feature Fusion module, termed ASCDC-CCFM, is designed in the neck to achieve efficient and lightweight fusion of multi-scale features from the backbone network. Furthermore, the Reparameterized Detection Head (RepHead) decouples classification and regression branches during training and reparameterizes them into a single convolution layer during inference, improving detection performance on degraded images, cluttered backgrounds, and dense small targets. Experiments conducted on the CUDD dataset show that LMOD-YOLO reduces computational cost and parameter count by 51.1% and 59.7%, respectively, and improves inference speed by 36.9% compared with YOLOv8m, with only slight decreases of 0.006 and 0.012 in mAP@0.5 and mAP@0.5:0.95.Deployment on an NVIDIA Jetson AGX Orin achieves a mean inference latency of 2.90 ms and an energy efficiency of 11.76 FPS/W. These results demonstrate a favorable balance between detection accuracy and deployment efficiency, enabling timely responses in dynamic underwater monitoring while potentially reducing the energy burden of battery-powered AUVs. The source code is available at https://github.com/yangxyyy/LMOD-YOLO.

An ultra-wide beam angle right-hand circularly polarized array antenna for UHF band is designed to improve the search and rescue coverage of emergency rescue UAVs effectively. The antenna is mainly composed of a radiating spiral arm board, a feed connection board and a power division phase-shifted feed network board, in which the radiating unit adopts a four-armed spiral antenna to reduce the size of the antenna array, achieving the right-hand circular polarization by controlling the phases and alignments of the excitation ports of the four spiral arms, adding an inverted F-branch between the radiating unit and the feed network to optimize the standing wave characteristics of the array, and implementing the feed network by a combination of multiple Wilkinson. The experimental results show that the array antenna has a voltage standing wave ratio (VSWR) < 2, half-power beamwidth (HPBW)>120 degrees, and gain> 4 dB in the 406-425 MHz band, where the above performances well satisfy the antenna requirements of the aviation emergency rescue system.

The traditional combination of navigation methods has been unable to meet all the needs of people in practical applic ations, and more and more researchers have put the research dir ection to computer vision, and the vision-based navigation metho ds have become a research hotspot, which can cover the anti-jam ming return of cluster UAVs, autonomous landing of UAVs and o ther application needs. Among them, the visual tracking technolo gy is of great significance for the navigation and landing of UAVs. In this paper, the research on parallel detection and tracking alg orithms is carried out on the airborne integrated control processi ng board composed of TMS320C6678 processor and Xilinx’s XC ZU7EV-2FFVC1156I MPSOC processor, and a parallel detection and tracking mechanism based on KCF is proposed, which solve s the scale problem of the KCF tracking algorithm; and its algorit hms are deployed in the aforementioned The algorithm is deploye d on the above mentioned airborne integrated control processing board and real-time optimization is carried out to achieve stable t racking of the target.

PCI Express (PCIe) is a high-speed serial computer expansion bus standard, designed to connect various hardware devices on the motherboard. It adopts a point-to-point connection and multi-lane architecture, offering a maximum transmission rate of up to $32 \text{GT} / \mathrm{s}$. It also supports hot-plugging and has low latency characteristics. Thanks to its high bandwidth and flexibility, PCIe is widely used in areas such as graphics cards, NVMe SSD, network adapters, and servers, making it a crucial connection technology in modern computer systems. This article presents a design of a PCIe Gen3 embedded switching system based on the PEX8748 chip. The system is verified through data exchange with five Yulong810A chips that integrate the PCIe Gen2 hard core. This enables the switching system to handle a large amount of data and device connections more efficiently and flexibly, thus adapting to the ever-changing technological requirements.

Current airborne microwave systems used for soil moisture detection are limited by their inability to achieve comprehensive real-time calibration, which consequently results in significant temperature-induced variations that affect detection accuracy. To address this challenge, the present study introduces the design of a high-efficiency L-band airborne dual-polarized microstrip antenna. By enhancing the antenna's radiation efficiency, this design effectively mitigates the influence of environmental temperature fluctuations on measurement outcomes, thereby improving detection precision. The antenna employs a dual-layer structure integrated with High Impedance Surface (HIS) technology, which successfully reduces return loss at the antenna ports, ensuring that the Voltage Standing Wave Ratio (VSWR) remains below 1.4. Experimental results obtained from an anechoic chamber demonstrate that, within the 1400-1430 MHz frequency range and at an elevation angle of 0 degrees, the horizontal polarization efficiency of the dual-layer dual-polarized HIS antenna achieves an average of 90.89%. Under identical testing conditions, its radiation efficiency is enhanced by 20% compared to that of a single-layer dual-polarized antenna.

To address feature degradation, scale sensitivity, and background interference in small object detection, we propose BOAD-YOLO-an enhanced YOLOv8s framework. Key innovations include: Backbone-integrated coordinate attention establishing long-range spatial dependencies; Neck with BiFPN for multi-scale feature fusion, DySample replacing standard downsampling to reduce computation, and nested coordinate attention enabling dual optimization; Head with $\mathbf{1 6 0} \times \mathbf{1 6 0}$ microdetection (P2) replacing $20 \times 20$ head (P5) and SCAM suppressing background interference via dual attention. Evaluations on VisDrone2019 show BOAD-YOLO achieves $\mathbf{4 5. 6 \%}$ mAP@0.5 ($+7.6 \%$ vs YOLOv8s) and $28.2 \% \mathrm{mAP} \text{@} 0.5: 0.95$ ($+5.1 \%$) with only 8.3 M parameters and 36.0 GFLOPs, providing an efficient solution for resource-constrained scenarios.

In order to solve the problem of trajectory capture and positioning of high-speed devices, as well as the high-performance requirements of multiple tasks such as real-time data acquisition and image storage processing when processing 1080P@ $\mathrm{5 0 0 F P S}$ high-speed video, this paper proposes a solution based on the XCZU7EV processor of the Zynq UltraScale+ MPSoC series as the hardware platform. The processor integrates a 4-core ARM Cortex-A53 PS and a programmable logic FPGA PL. The FPGA is designed to solve the problem of real-time high-speed image acquisition. The 4 -core ARM is used as the system master to solve the multi-task processing scheduling. Each core can communicate with the FPGA through the internal AXI interconnect bus. The main control system uses the concurrent network software framework to interact with the host client through the Ethernet interface. TCP network services are performed, and the multi-core ARM processor is combined for task scheduling, realizing the extraction and processing of highframe rate images, video output display in SDI format, and storage management and retrieval and upload of high-speed data ECM files. The key functions of multi-task simultaneous execution and efficient event response have been realized, which significantly improves the overall integration and performance of the system.

In recent years, the MLVDS (Multi Point Low Voltage Differential Signaling) standard has been widely used in serial interfaces. FPGAs are commonly used as transmission and reception chips for MLVDS signals. Compared with the traditional MLVDS protocol based on standard protocols such as UART and SPI, this paper proposes a dual redundant fast response MLVDS system scheme based on FPGA (Field Programmable Gate Array) platform. The half duplex characteristic of the MLVDS standard makes it difficult for the system to switch between transmission and reception in a timely manner, thereby reducing the transmission rate. This design solves this problem by customizing protocols and completing protocol parsing and packet grouping within FPGA. Meanwhile, the design adopts dual redundancy switching to ensure the stability and reliability of the high-speed communication system.

Facing the real-time processing and high reliability requirements of the intelligent processing load on board, a real-time detection and identification system for on orbit satellites is designed based on the hardware platform composed of Xilinx V7 series FPGA 690T FFG1158-2 and Zhuhai obit Yulong Series CPU Yulong810a. The FPGA in the system receives and forwards the load data, and the Yulong810a in the system runs the neural network to detect, identify and locate the target of the data forwarded by the FPGA. Several state machines are added to the FPGA side to realize real-time state monitoring and multiple handshake protection mechanism, and the process of software annotation of onboard load is optimized. After the test and verification of the ground inspection equipment, the system can meet the high real-time requirements of the on-board system, complete the real-time detection and recognition task and return the target information to the ground.

In recent years, with the increasing tensions in international affairs and the intensification of territorial disputes, the development and protection of marine resources in the South China Sea and the East China Sea have faced significant challenges. The increasingly frequent military activities of neighboring countries have further complicated the maritime security situation. Horizon line detection plays a crucial role in marine engineering and security defense. However, in real-world maritime environments, horizon detection is often hindered by factors such as clouds, waves, fog, islands, and coastlines.To accurately detect the horizon in actual marine conditions, using long-wave infrared images collected from both simulations and real-world environments, this paper proposes a scene-adaptive multi-feature and multi-classification scene classification method for sensing sea-surface interference types. By categorizing the interference types based on how various factors in different scenes affect horizon detection, and by designing corresponding improvements to horizon detection methods based on spectral characteristics, the proposed approach effectively filters out noise and disturbances, precisely distinguishing the horizon. This enhances the accuracy and reliability of sea-surface scene perception and analysis.

Classification of transparent materials with various roughness types has been widely used in the field of computer vision. However, the surface roughness of a transparent material affects the extraction effect of classification features, thus affecting the performance of transparent material classification. In this study, a classification method of transparent materials with various surface roughness types and transparencies, which uses the microfacet shape factor, reflectivity, and transmissivity as classification characteristics, is proposed. First, a transparent material feature extraction method based on microfacet distribution function is proposed for the first time, and the microfacet shape factor, reflectivity, and transmissivity are extracted by our model as classification features. The microfacet distribution model ground glass unknown is combined with the time-of-flight imaging model to achieve an accurate classification of surfaces with various roughness types. Then, according to the nonlinear and discrete characteristics of data, an appropriate classifier is selected to realize the transparent material classification. The transparent material classification experiments are performed using four types of material appearances, and the proposed method is compared with the methods of Shim et al. and Lang et al. The average classification accuracy of the proposed method for the transparent materials with four material appearances is 92.62

In recent years, LVDS transmission has been widely used in high-speed information transmission [1], and FPGA is used as a receiver chip for LVDS signals in high-speed data transmission examples. Compared with the traditional LVDS data interaction based on SRIO [2], AROURA [3] and other protocols, this paper designs a data processing scheme for high-speed CMOS under the FPGA platform, high-speed CMOS data output without coding, multi-channel (from tens to hundreds of channels) parallel transmission form, so that the timing control and data restoration of the acquisition end is more difficult, this design through the multi-channel LVDS data bit reception, parallel processing and channel synchronization work. The design of the whole set of data acquisition scheme is completed, and the method is verified by experiments, which realizes the reception, parallel processing and channel synchronization of LVDS data in each channel, ensuring the stability and reliability of data during transmission.

Aiming at the problems of low efficiency of feature extraction and insufficient use of feature correlation in image retrieval based on deep hashing, an image retrieval method Dual Learning Hashing(DLH) that combines feature layer learning and hash layer learning is proposed. In view of the problem that existing methods usually extract local features with intensive attention mechanism by focusing on dense local regions, these local regions cannot contain different local information, a module is developed to learn differentiated local features by locating the peaks of non-overlapping subdomains in the feature map. In order to make full use of semantic information and generate high quality hash code, quantization function and probabilistic semantic-preserving function are designed in the hash layer. Finally, combined with the two parts of learning, the total loss function is given. A large number of comparative experiments have been carried out on three widely used datasets. The experimental results show that compared with other advanced deep hash methods, DLH has achieved better retrieval performance.

The recognition of airport apron targets and the determination of target positions are critically important in positioning the apron when there are abnormal target positions or unmanned aerial vehicles (UAVs) engaged in autonomous cargo retrieval. This is crucial for ensuring the safety and efficiency of the apron. However, challenges such as resource-intensive processes and suboptimal accuracy in position solving have been encountered due to limitations in hardware equipment and the complexity of multi-sensor fusion. In this paper, we propose an adaptive target location method for airport aprons based on monocular high-speed cameras. This method employs adaptive frame differencing and perspective transformation based on geometric shape extraction to detect apron objects and calculate their azimuth and deflection from the apron center. Importantly, our approach relies solely on a single visible-light camera, eliminating the need for complex multi-sensor fusion involving binocular matching, Global Navigation Satellite Systems (GNSS), or Inertial Measurement Units (IMU). Experimental results demonstrate that the proposed method achieves position solving accuracy within 1 meter and real-time recognition and solving capabilities for apron objects.

Fault injection technology is an important part of embedded system testing, it can simulate the software anomaly and hardware failure in the system. Fault injection testing has become a part of many device tests. This paper provides a hardware design based on ZYNQ 8-channel input-output module with fault injection. Each channel can support analog signal acquisition, analog signal output, and fault injection function for analog signal output. The channel type and fault injection type are determined through software configuration, which greatly improves the flexibility of the device.

With the development of technology, image and other data are developing in the direction of higher resolution and higher frame rate, which puts forward higher requirements for the speed of big data real-time storage systems. At the same time, in order to facilitate the use of outdoor and other practical scenes, more miniaturization and integration of storage system are required.In order to solve the problems of high-speed real-time storage, device integration and miniaturization, this paper designs an NVME SSD real-time high-speed storage system based on ZYNQMP SoC. The ZYNQMP SoC is connected to the NVME SSD through PCIE interface, and the NVME protocol is migrated to complete the interaction with the NVME SSD, so as to achieve higher rate of read and write. Migrate the exFAT file system to manage files in NVME SSD.The results show that the system can read and write NVME SSD at a speed of 1600MB/s. After NVME SSD is inserted into a PC, exFAT file system can be recognized and files can be read and written normally, and real-time, high-speed and stable storage is realized.

The classification of materials is a research hotspot. These methods generally focus on the classification of flat materials and do not consider the influence of polishing and convex surfaces. We develop a classification algorithm of polishing and convex surface objects, and derive the photon accumulation point spread function (PAPSF) of material from the imaging model of a binocular pulsed time-of-flight (ToF) camera as the classification feature, which consists of depth distortion, the indirect reflection photon cumulant and the indirect reflection photon cumulant. We design a one-versus-all support vector machine (SVM) classifier to classify materials of polishing and convex surfaces objects. We conduct classification experiments on four plastics and four metal materials with a similar appearance. Our method in flat and raw material classification has the same classification accuracy as the latest method based on a continuous-wave- modulation ToF camera, but also our method achieved accuracies of 91.0% in flat and polishing material classification, 93.0% in different convex surface and fixed polishing material classification, 91.5% in fixed convex surface and different polishing material classification and 90.2% in polishing and convex surface material classification.

This paper is based on the Zynq UltraScale plus EV series MPSoc hardware platform. To meet the demand of 1080P@500FPS high-speed video with large capacity data storage and high performance, multiple Cortex-A53 cores on the ARM side of the processor are implemented with shared memory, inter-core interrupt mechanism, and multi-core supervised inter-core communication mechanism. The system's multi-task processing and parallel control are realized through multi-core scheduling, which effectively improves the system's high integration, flexibility, and reliability.