Infrared noncontact detection has the advantages of having no power cut-off and convenient operation. However, when the SF ₆ (sulfur hexafluoride) gas leakage is small, the existing detection algorithms have low accuracy and are easily disturbed by noise. To solve this problem, this article proposes an online detection method for SF ₆ gas leakage based on a Gaussian mixture model. First, as a preprocessing stage, the improved dynamic interframe time domain filter is used to suppress the random noise in the image. Then, contrast limited adaptive histogram equalization (CLAHE) is used to enhance the dark details of the infrared image to improve the local contrast. Finally, the improved Gaussian mixture background model is used to adaptively segment the SF ₆ gas leakage and mark the leakage area. The results show that under the experimental conditions of 0.06 mL/min indoor and 5-m distance, the algorithm can overcome the high noise and complex background disturbances of infrared imaging. It can effectively detect and locate the SF ₆ leakage area and gets a higher F1-score compared with other conventional algorithm under the experimental conditions. In addition, when the resolution of the infrared image is 320 ×240, the running efficiency of the algorithm meets the real-time requirements.

Detection tasks based on supervisory control and data acquisition (SCADA) data are challenging because of the following problems: (i) high redundancy and high non-linearity of data complicate the detection task; (ii) because the sheer volume of data, the current methods are hard to accurately detect useful early fault information from raw data. This study proposes a wind turbine blade cracking early fault detection model, the parallel multiple CNNs with temporal predictions (PMCTP). The model innovatively integrates temporal convolutional network (TCN), ResNet50, and Xception. Specifically, PMCTP applies the TCN model to SCADA data in a parallel branch to improve detection accuracy by extracting temporal features predicted by the TCN model in parallel. The PMCTP reduces the impact of data redundancy and high data volumes by dividing the wind turbine data into four categories according to different state parameters. In addition, PMCTP combines the advantages of ResNet50 and Xception to perform adaptive feature extraction for each of the four types of state parameter data for a wind turbine. PMCTP introduces multiple fully connected layers into the model, which can provide parallel computation for serializing hidden features. Experimental results show that PMCTP is more accurate than some popular convolutional neural networks.

Dynamic environments are challenging for visual Simultaneous Localization and Mapping, as dynamic elements can disrupt the camera pose estimation and thus reduce the reconstructed map accuracy. To solve this problem, this study proposes an approach for eliminating dynamic elements and reconstructing static background in indoor dynamic environments. To check out dynamic elements, the geometric residual is exploited, and the static background is obtained after removing the dynamic elements and repairing images. The camera pose is estimated based on the static background. Keyframes are then selected using randomized ferns, and loop closure detection and relocalization are performed according to the keyframes set. Finally, the 3D scene is reconstructed. The proposed method is tested on the TUM and BONN datasets, and the map reconstruction accuracy is experimentally demonstrated.

Infrared and visible images of the same scene are fused to produce a fused image with richer information. However, most current image-fusion algorithms suffer from insufficient edge information retention, weak feature representation, and poor contrast, halos, and artifacts, and can only be applied to a single scene. To address these issues, we propose a novel infrared and visual image fusion algorithm based on a bilateral–least-squares hybrid filter (DBLSF) with the least-squares and bilateral filter hybrid model (BLF-LS). The proposed algorithm utilizes the residual network ResNet50 and the adaptive fusion strategy of the structure tensor to fuse the base and detail layers of the filter decomposition, respectively. Experiments on 32 sets of images from the TNO image-fusion dataset show that, although our fusion algorithm sacrifices overall time efficiency, the Combination 1 approach can better preserve image edge information and image integrity; reduce the loss of source image features; suppress artifacts and halos; and compare favorably with other algorithms in terms of structural similarity, feature similarity, multiscale structural similarity, root mean square error, peak signal-to-noise ratio, and correlation coefficient by at least 2.71%, 1.86%, 0.09%, 0.46%, 0.24%, and 0.07%; and the proposed Combination 2 can effectively improve the contrast and edge features of the fused image and enrich the image detail information, with an average improvement of 37.42%, 26.40%, and 26.60% in the three metrics of average gradient, edge intensity, and spatial frequency compared with other algorithms.

An accurate close-range positioning and navigation system is the key to autonomous cruise and the fine operations of multi-rotor UAVs. In this paper, a dynamic electromagnetic positioning system is designed and implemented to accurately navigate a UAV to a very small landing zone. The system is composed of three quasi-static electromagnetic field generators (EGs) installed around the landing zone, and one electromagnetic field sensor (ES) mounted on the UAV. The ES adaptively controls its gain and adjusts the power of the EGs, so it can detect electromagnetic signals under a high SNR (signal-to-noise ratio) condition and has an anti-saturation ability. Consequently, the coordinate and azimuthal angle of the UAV can be measured during the process of the UAV approaching the landing zone. Hardware experiments demonstrate that the prototype system can effectively navigate a UAV to a small landing zone that is slightly larger than the size of the UAV, and the terminal positioning error is less than 2 cm.

A new image registration method for infrared images and visible images is proposed in this paper. This method is a modified version of SIFT algorithm. Salient points of infrared images and visible images are extracted along edges of images. Each salient point is assigned a dominant direction which is the peak of orientation histogram of local image around it and a feature vector which is formed from the orientation histograms of sub-region around it. After salient points matching, RANSAC algorithm is used to eliminate wrong corresponding pairs. Experimental results show that this method has a good registration result for infrared images and visible images.