Neuromorphic hardware systems have been gaining ever-increasing focus in many embedded applications as they use a brain-inspired, energy-efficient spiking neural network (SNN) model that closely mimics the human cortex mechanism by communicating and processing sensory information via spatiotemporally sparse spikes. In this paper, we fully leverage the characteristics of spiking convolution neural network (SCNN), and propose a scalable, cost-efficient, and high-speed VLSI architecture to accelerate deep SCNN inference for real-time low-cost embedded scenarios. We leverage the snapshot of binary spike maps at each time-step, to decompose the SCNN operations into a series of regular and simple time-step CNN-like processing to reduce hardware resource consumption. Moreover, our hardware architecture achieves high throughput by employing a pixel stream processing mechanism and fine-grained data pipelines. Our Zynq-7045 FPGA prototype reached a high processing speed of 1250 frames/s and high recognition accuracies on the MNIST and Fashion-MNIST image datasets, demonstrating the plausibility of our SCNN hardware architecture for many embedded applications.

The construction of smart courts promotes the in-deep integration of internet, big data, cloud computing and artificial intelligence with judicial trial work, which can both improve trials and ensure judicial justice with more efficiency. High-quality structured legal facts, obtained by extracting information from unstructured legal texts, are the foundation for the construction of smart courts. Based on the strong normative characteristics of Chinese legal text content and structure composition and the strong text feature learning ability of deep learning, this paper proposes an ontology-based and deep learning-driven method for extracting legal facts from Chinese legal texts. The proposed method utilizes rules and patterns generated in the process of knowledge modeling to extract simple entities, and then extracts complex entities hidden in legal text details with deep learning methods. Finally, the extracted entities are mapped into structured legal facts with clear logical relationships by the Chinese Legal Text Ontology. In the information extraction test of judicial datasets composed of Chinese legal texts on theft, the proposed method effectively extracts up to 38 categories of legal facts from legal texts and the number of categories extracted increases significantly. Among them, the rule-based extractor obtains an F1-score of 99.70%, and the deep learning-driven extractor obtains an F1-score of 91.43%. Compared with existing methods, the proposed method has great advantages in extracting the completeness and accuracy of legal facts.

Taking advantage of the functional complementarity between infrared and visible light sensors imaging, pixel-level real-time image fusion based on infrared and visible light images of different resolutions is a promising strategy for visual enhancement, which has demonstrated tremendous potential for autonomous driving, military reconnaissance, video surveillance, etc. Great progress has been made in this field in recent years, but the fusion speed and quality of visual enhancement are still not satisfactory. Herein, we propose a multi-scale FPGA-based image fusion technology with substantially enhanced visual enhancement capability and fusion speed. Specifically, the source images are first decomposed into three distinct layers using guided filter and saliency detection, which are the detail layer, saliency layer and background layer. Fusion weight map of the saliency layer is subsequently constructed using attention mechanism. Afterwards weight fusion strategy is used for saliency layer fusion and detail layer fusion, while weight average fusion strategy is used for the background layer fusion, followed by the incorporation of image enhancement technology to improve the fused image contrast. Finally, high-level synthesis tool is used to design the hardware circuit. The method in the present study is thoroughly tested on XCZU15EG board, which could not only effectively improve the image enhancement capability in glare and smoke environments, but also achieve fast real-time image fusion with 55FPS for infrared and visible images with a resolution of 640 × 470.

Fluorescence images enhancement is important as it can provide more information for medical diagnosis. In this work, we design three simple yet useful filters based on the combinations of mathematical functions, which are proved to be effective in strengthening the images acquired from the fluorescence microscope. Using these filters, detailed objects can be found in the dark sections of the fluorescence images. In addition, these filters can be used to enhance the low-light image, which provide satisfactory visual information and marginal profile for the blurred objects in the image. Moreover, these filters have been used to enhance the image with high degradation by the Gaussian noise, where clear edge profile can be extracted. Finally, we have shown that these filters can be utilized for the image compression. Compression ratio can be obtained to be 0.9688. This study shows the making of the filters with dual functions for the image enhancement and the image compression. Our designed filters are showing the potentials in the field of biomedical imaging and pattern identification.

Temperature-adjustable optical properties of two types of one-dimensional photonic superlattices made from lithium niobate or liquid crystal are theoretically investigated. The control of band structures and density of states is achieved through the temperature with various values of refractive index, layer width, and propagation wavelength. It is presented that a null photonic gap appears at certain temperature points, which are decided by the ratio of layer width and the refractive index. It is shown that the spatial average of the wave vector ⟨ k ⟩ vanishes at a specific temperature as functions of the layer thickness and refractive index. The values of the temperature corresponding to zero-refractive-index (zero- ⟨ n ⟩ ) are demonstrated to be suppressed when the refractive index or the ratio of the layer width is enhanced. Null photonic gap and zero- ⟨ n ⟩ can appear simultaneously at certain wavelengths in the range of visible light, which can be modulated by the temperature. This implies the importance of one-dimensional photonic superlattices made from temperature-controllable-refractive-index materials for many important practical applications.

A cyclic redundancy check (CRC) is a widely used technique in data communication for detecting data transmission errors. However, existing FPGA-based parallel CRC hardware implementation schemes often face problems of excessive resource utilization and timing convergence difficulties in high-bandwidth networks. In addition, these problems are further exacerbated by the variable length of the end of the checksum data frame during data transmission. To address these challenges, this paper proposes a parallel CRC computation method based on precomputed seed values for bit-width normalization (named PSV-WN-CRC). The algorithm selects the corresponding primitive seed value according to the length of the data frame tail and converts the CRC computation with arbitrary bit-width to the CRC computation with fixed bit-width, thus adapting to the case of the indefinite length of the data frame tail. Based on this algorithm, this paper designs an efficient parallel CRC circuit on FPGA to reduce the consumption of resources. The experimental results show that the CRC algorithm and circuit proposed in this paper implemented on Virtex UltraScale+ FPGAs with 1024-bit wide CRC consumes only 5981 LUTs and achieves a maximum throughput of 392.2 Gbps. The method effectively reduces the resource consumption and improves the maximum throughput as compared to three advanced works.