Deep learning-based video watermarking algorithms perform well in terms of robustness and perceptual quality. However, their resistance to HEVC compression remains a major limitation, especially under high compression ratios, where watermark extraction accuracy significantly degrades. To address this issue, this paper proposes a Spatio-temporally Enhanced Video Watermarking (SEVMark) based on invertible neural networks (INNs). SEVMark introduces a channel attention mechanism in the temporal domain to adaptively focus on keyframes, and employs spatial pyramid pooling module in the spatial domain to capture multi-scale features. These two modules work in tandem to enhance the spatio-temporal feature representation, achieving high robustness and imperceptibility. Furthermore, based on the HEVC encoding process, a HEVC video compression simulator (DiffH265) is designed and incorporated as a key component of the noise layer, guiding the encoder-decoder network to maintain high extraction accuracy under HEVC compression. Experimental results demonstrate that SEVMark outperforms state-of-the-art methods in both quantitative and qualitative evaluations, particularly demonstrating excellent robustness against HEVC compression attacks under high compression ratios.

With high embedding capacity and security, transform coefficient-based video steganography has become an important branch of video steganography. However, existing steganalysis methods against transform coefficient-based steganography provide insufficient consideration to the prediction process of HEVC compression, which results in steganalysis that is not straightforward and fail to effectively detect adaptive steganography methods in low embedding rate scenarios. In this paper, an HEVC video steganalysis method based on centralized error and attention mechanism against transform coefficient-based steganography is proposed. Firstly, the centralized error phenomenon brought by distortion compensation-based steganography is analyzed, and prediction error maps is constructed for steganalysis to achieve higher SNR(signal-to-noise ratio). Secondly, a video steganalysis network called CESNet (Centralized Error Steganalysis Network) is proposed. The network takes the prediction error maps as input and four types of convolutional modules are designed to adapt to different stages of feature extraction. To address the intra-frame sparsity of adaptive steganography, CEA (Centralized Error Attention) modules based on spatial and channel attention mechanisms are proposed to adaptively enhance the steganographic region. Finally, after extracting the feature vectors of each frame, the detection of steganographic video is completed using the self-attention mechanism. Experimental results show that compared with the existing transform coefficient-based video steganalysis methods, the proposed method can effectively detect multiple transform coefficient-based steganography algorithms and achieve higher detection performance in low payload scenarios.

HEVC (High Efficiency Video Coding) provides abundant embedding carriers for video steganography, leading to rapid development in the field of video steganography while increasing the urgent demand for video steganalysis. However, existing steganalysis methods against PU (prediction unit) based steganography primarily use the extraction of video statistical features, which ignore the potential information of each frame and fail to effectively detect different PU-based steganography methods. In this paper, a video steganalysis method based on PU maps and multi-scale convolutional residual network is proposed. Firstly, the effects of PU-based steganography on the spatial domain and the compressed domain are analyzed. It is observed that steganography has less impact on the spatial domain, whereas it significantly disrupts the connection between PU blocks in the compressed domain, leaving distinct steganographic traces. Consequently, the PU partition modes containing local connections are introduced to generate PU maps for steganalysis. Secondly, a video steganalysis network called PUSN (Prediction Unit Steganalysis Network) is constructed. The network takes PU maps as input and consists of three parts: feature extraction, feature representation, and binary classification. Additionally, a multi-scale module is proposed to enhance the detection performance. Finally, the detection result of the steganographic video is obtained by the voting mechanism. The experimental results show that compared with the existing steganalysis methods, the proposed method could effectively detect multiple PU-based steganography methods and achieve higher detection accuracy across various embedding rates.

In recent years, with the rapid development of tourism, the abnormal events in the traffic fortress of tourist attractions not only endanger personal safety, but also cause a lot of negative effects on the society. In the face of the low accuracy of abnormal event detection in complex scenes and the low efficiency of manual observation, combining the advantages and characteristics of YOLOv5 and 3D convolutional neural network(C3D), an automatic abnormal event detection algorithm based on YOLOv5+C3D was proposed. Experimental data show that compared with other methods, the abnormal event detection method based on YOLOv5+C3D has a higher accuracy for various abnormal events detection, indicating that the trained model has a strong generalization ability for abnormal event detection.

In order to solve the problems of low signal-to-noise ratio, uncertain position, and blurred edges of mammography images, and to achieve full-view breast mass segmentation, a breast mass segmentation network based on improved UNet 3+ is proposed. The network provides prior knowledge of edge information for model segmentation by adding an edge-aware module and utilizing shallow and deep features in the network. At the same time, Hybrid loss function (HAM) is added to the network to obtain rich multi-scale information for handling lumps of different sizes and shapes. In addition, we add Dice loss and bce loss to the hybrid loss function to address the problem of pixel class imbalance in mammography images. The experimental results show that the improved UNet 3+ segmentation model has reached the Dice scores of 85.89% and 82.55% on the CBIS-DDSM and INbreast datasets, has improved aa and bb compared with the previous ones, and has a higher performance in the breast mass segmentation task. The accuracy rate is better than other classical models.

A lightweight object detection algorithm based on YOLOv5 is proposed to address the problem of deploying detection models for traffic targets. This method was proposed to reduce the number of channels in the backbone and introduced GSConv to improve performance. At the same time, GSConv was improved to cut the parameters. The C3 module was replaced in the Neck with C2f to obtain more comprehensive gradient flow information. Finally, the model parameters are only 1.41M. Experiment on BDD100K public traffic dataset shows that the lightweight model reduces the number of parameters while losing a small amount of Recalls, and its performance is better than mainstream lightweight networks.

In certain industries such as construction, high risks are often associated with the construction process, and safety helmets are crucial protective gear for workers on construction sites. To address the issues of missed and false detections of safety helmets in complex environments with current helmet detection methods, we propose an improved YOLOv5 object detection algorithm to detect the wearing of safety helmets. The proposed improvements include the addition of an attention mechanism and replacement of the loss function. The proposed method adds an efficient channel attention (ECA) mechanism to the YOLOv5 head network to enhance the model's ability to extract features related to safety helmets, thereby increasing precision without adding too much computational complexity. The loss function is replaced with NWD to effectively improve the detection progress of small safety helmets in YOLOv5.