As one of the important features in the early stage of fires, the detection of smoke can provide a faster early warning of a fire, thus suppressing the spread of the fire in time. However, the features of smoke are not apparent; the shape of smoke is not fixed, and it is easy to be confused with the background outdoors, which leads to difficulties in detecting smoke. Therefore, this study proposes a model called Smoke Detection Transformer (Smoke-DETR) for smoke detection, which is based on a Real-Time Detection Transformer (RT-DETR). Considering the limited computational resources of smoke detection devices, Enhanced Channel-wise Partial Convolution (ECPConv) is introduced to reduce the number of parameters and the amount of computation. This approach improves Partial Convolution (PConv) by using a selection strategy that selects channels containing more information for each convolution, thereby increasing the network’s ability to learn smoke features. To cope with smoke images with inconspicuous features and irregular shapes, the Efficient Multi-Scale Attention (EMA) module is used to strengthen the feature extraction capability of the backbone network. Additionally, in order to overcome the problem of smoke being easily confused with the background, the Multi-Scale Foreground-Focus Fusion Pyramid Network (MFFPN) is designed to strengthen the model’s attention to the foreground of images, which improves the accuracy of detection in situations where smoke is not well differentiated from the background. Experimental results demonstrate that Smoke-DETR has achieved significant improvements in smoke detection. In the self-building dataset, compared to RT-DETR, Smoke-DETR achieves a Precision that has reached 86.2%, marking an increase of 3.6 percentage points. Similarly, Recall has achieved 80%, showing an improvement of 3.6 percentage points. In terms of mAP50, it has reached 86.2%, with a 3.8 percentage point increase. Furthermore, mAP50 has reached 53.9%, representing a 3.6 percentage point increase.

Unmanned aerial vehicle (UAV) and base station jointly assisted multi-access edge computing (UB-MEC) technology is a promising direction to provide flexible computing services for resource-limited devices. Due to the non-real-time observation of device loads and the dynamic nature of demand in UB-MEC, it is a highly challenging problem to make UAV respond in real time to meet user’s dynamic preferences in UB-MEC. To this end, we propose a multi-objective deep reinforcement learning (MODRL) for computation offloading and trajectory control (COTC) of UAV. First, the problem is formulated as a multi-objective Markov decision process (MOMDP), where the traditional scalar rewards are extended to vector, corresponding to the number of task data collected, the completion delay and the UAV’s energy consumption, and the weights are dynamically adjusted to meet different user preferences. Then, considering the device load information stored in UAV is non-real-time, an attentional long short-term memory (ALSTM) network is designed to predict real-time states by autofocusing important historical information. The near on-policy experience replay (NOER) reviews experiences close to on-policy can better promote learning of current strategy. The simulation results show that the proposed algorithm can obtain the action policy which meets the user’s time-varying preferences, and can achieve a good balance between different objectives under different preferences.

Accurate and efficient detection of floating waste is crucial for environmental protection and aquatic ecosystem preservation, yet remains challenging due to environmental interference and the prevalence of small targets. To address these limitations, we propose a Multi-scale Adaptive Real-time Detector (RTDETR-MARD) based on RT-DETR that introduces three key innovations for improved floating waste detection using unmanned surface vessels (USVs). First, our hierarchical multi-scale feature integration leverages the gather-and-distribute mechanism to enhance feature aggregation and cross-layer interaction. Second, we develop an advanced feature fusion module incorporating feature alignment, Information Fusion, information injection, and Scale Sequence Feature Fusion components to ensure precise spatial alignment and semantic consistency. Third, we implement the Wise-IoU loss function to optimize localization accuracy through high-quality anchor supervision. Extensive experiments demonstrate the framework’s effectiveness, achieving state-of-the-art performance of 86.6% mAP50 at 96.8 FPS on the FloW dataset and 49.2% mAP50 at 107.5 FPS on our custom water surface waste dataset. These results confirm RTDETR-MARD’s superior accuracy, real-time capability, and robustness across diverse environmental conditions, making it particularly suitable for practical deployment in ecological monitoring systems where both speed and precision are critical requirements.

Fire is a common but serious disaster, which poses a great threat to human life and property. Therefore, fire-smoke detection technology is of great significance in various fields. In order to improve the detection ability of tiny-fire, so as to realize the prediction and suppression of fire as soon as possible, we proposed an efficient and accurate tiny-fire detection method based on the optimized YOLOv7, and we named the improved model YOLOv7-FIRE. First, we introduced the BiFormer into YOLOv7 to make the network pay more attention to the fire-smoke area. Second, we introduced the NWD technique to enhance the perception of the algorithm for small targets, and provided richer semantic information by modeling the context information around the target. Finally, CARAFE was applied for content-aware feature reorganization, which preserved the details and texture information in the image and improved the quality of fire-smoke detection. Furthermore, in order to improve the robustness of the improved algorithm, we expanded the fire-smoke dataset. The experimental results showed that YOLOv7-FIRE as significantly better than the previous algorithm in detection accuracy and recall rate, the Precision increased from 75.83% to 82.31%, and the Recall increased from 66.43% to 74.02%.