With the advancement of intelligent monitoring and computer vision, real-time monitoring systems integrating mobile edge devices and cloud computing are increasingly used in safety management of crowded scenarios. This paper proposes BCLD-YOLOv11, a real-time pedestrian detection model for crowded scenarios. Firstly, a C3k2_EMBConv module is proposed. This module introduces an enhanced EMBConv based on the eSE attention mechanism, effectively improving the model’s ability to recover details of occluded pedestrians. Secondly, a lightweight feature fusion network, LDEFPN, is designed. This network enhances the fusion of multi-scale pedestrian features through the Enhance_FPN module. It adopts DySample as the upsampling module to reduce computational burden while maintaining efficient feature representation, and utilizes the AEP path to enhance the semantic expression and spatial details of high-level features without increasing computational cost, further improving the model’s perception of multi-scale and occluded pedestrians. Experimental results show that BCLD-YOLOv11 achieves a real-time detection speed of 192.31 FPS. Compared to YOLOv11n, it reduces Parameters by 18.5%, decreases Model Size from 5.23M to 4.45M, and improves mAP@50 by 1.68%, meeting real-time detection requirements for pedestrians in crowded scenarios.

Underwater images often suffer from image quality degradation due to the influence of refraction and reflection from suspended matter in water, as well as uneven light absorption. This can result in image blurring, loss of detail, fogging, and color distortion, which can significantly impact underwater visual tasks. This paper proposes a method for enhancing the quality of underwater images through multi-scale attention hybrid feature fusion. The input underwater image undergoes multiscale feature extraction using a multiscale attention hybrid feature extraction block to obtain different levels of feature representations. Feature fusion and enhancement are then achieved through a multilevel feature depth coupling module using multiple Resize Blocks to couple the multiscale features in depth. The combination of these two modules effectively solves the problem of degraded picture quality in underwater images. Experimental results demonstrate that the proposed method in this paper improves PSNR by 7.897%, SSIM by 1.724%, and VIF by 2.796%, proving the algorithm’s effectiveness and superiority.

Face tampering is an intriguing task in video/image genuineness identification and has attracted significant amounts of attention in recent years. In this work, we propose a face forgery detection method that consists of preprocessing, an improved Siamese network-based feature extractor (including a feature alignment module), and postprocessing (a voting principle). Roughly speaking, our method extracts the features in the grey space of face/background image pairs and measures the difference to make decisions. Experiments on several standard databases prove the effectiveness of our method, and especially on the low-quality subdataset of the FaceForensics++ , our method achieves a competitive result.

In the process of research and development of speech recognize interface for Z+Z intelligence education platform, there are too many component interfaces and their usage is too difficult. A strategy for designing software, named re-integrating of architecture/component oriented general model of sub-domain, is proposed on the basis of domain-specific application system, in order to simplify architectures and system integration. Feasibility and design method are demonstrated, and a general component for recognizing voice-controlled command is designed. The DMDSS (Development Model of Domain-Specific Software) based on component/architecture is improved by Yajie and Junqiao, (2003). This idea of software design can be complemented of DMDSS.