Autonomous driving (AD) systems requires multisensor collaboration to address challenges caused by complex weather scenarios. The recognition accuracy of autonomous driving system based on single resource, either on image only or Lidar only, becomes unreliable due to untested weather conditions, occlusions objects, and other factors. This paper proposes a decision-level fusion network based on an improved YOLOV7 and an improved CenterPoint network to build a multi-sensor-collaboration scheme. The overall accuracy of the proposed fusion algorithm is improved for small targets by adding multi-scale and multi-stage attention channel modules into the backbones of image recognition network and point cloud recognition network respectively. Moreover, the fusion algorithm introduces mixed distance constraints as the loss function for overlapping targets. The proposed fusion algorithm has been successfully tested on the public ONCE dataset mixed with a self-built dataset under various road conditions such as sunny, night, and rainy weather. The mAP of proposed decision-level fusion algorithm achieves$\mathbf{8 3. 5 \%}$in sunny daytime,$\mathbf{8 0 \%}$during nighttime and 79.1 % during rain time.

Scene perception for autonomous vehicles and vessels is crucial for autonomous navigation. Current mainstream transformer methods typically split the feature map into windows, such as local, dilated, and horizontal/vertical bar windows. However, their token interaction is confined to fixed windows, posing challenges for image-based semantic segmentation. This article proposes a novel model, LBFormer, which enables flexible token interaction across different windows. Specifically, a window-level affinity graph is constructed from coarse-grained features using self-attention clustering and evolves during training, retaining top-k windows with high semantic relevance for each window. Self-attention purification is then employed to compress and filter fine-grained features with low semantic relevance within the top-k windows, ensuring effective token interaction for each feature point. To enhance context modeling within windows and build a more effective window-level affinity graph, a dual branch method extracts multidimensional features from each window, which are then interacted with and fused via the feature aggregation module. Extensive experiments at an image resolution of 224×224 were conducted on our private YZ-DATA water surface scene dataset and the public CamVid urban scene dataset. The results show that LBFormer achieves an MIoU of 89.80% on YZ-DATA and 61.31% on CamVid, surpassing mainstream transformer methods.

Video Anomaly Detection (VAD) remains a fundamental yet formidable task in the video understanding community, with promising applications in areas such as information forensics and public safety protection. Due to the rarity and diversity of anomalies, existing methods only use easily collected regular events to model the inherent normality of normal spatial-temporal patterns in an unsupervised manner. Although such methods have made significant progress benefiting from the development of deep learning, they attempt to model the statistical dependency between observable videos and semantic labels, which is a crude description of normality and lacks a systematic exploration of its underlying causal relationships. Previous studies have shown that existing unsupervised VAD models are incapable of label-independent data offsets (e.g., scene changes) in real-world scenarios and may fail to respond to light anomalies due to the overgeneralization of deep neural networks. Inspired by causality learning, we argue that there exist causal factors that can adequately generalize the prototypical patterns of regular events and present significant deviations when anomalous instances occur. In this regard, we propose Causal Representation Consistency Learning (CRCL) to implicitly mine potential scene-robust causal variable in unsupervised video normality learning. Specifically, building on the structural causal models, we propose scene-debiasing learning and causality-inspired normality learning to strip away entangled scene bias in deep representations and learn causal video normality, respectively. Extensive experiments on benchmarks validate the superiority of our method over conventional deep representation learning. Moreover, ablation studies and extension validation show that the CRCL can cope with label-independent biases in multi-scene settings and maintain stable performance with only limited training data available.

Autonomous driving technology has become increasingly important in recent years, with the potential to revolutionize transportation systems and improve road safety. Vision-based methods have long been used in this field, but the major challenges in object detection are efficiency and occlusion. To address this challenge, anchor-free collaborative detection has been proposed as a promising solution. Despite its potential, there has been limited research on this approach. This study proposes an efficient vision-based multi-view object detection and localization method that leverages anchor-free collaborative detection to improve the accuracy of pedestrian detection. The method first generates feature maps to extract the head and foot of pedestrians and then applies spatial aggregation to fuse information from different views. Additionally, the study examines the efficiency of different convolutional neural network architectures for the feature map extraction model and identifies ResNet18 and ResNet34 as the most efficient models for the task. The proposed method has the potential to significantly improve the accuracy of pedestrian detection and localization in autonomous driving scenarios, which is critical for ensuring safety. Overall, this work contributes to the development of vision-based methods for autonomous driving and has significant implications for the future of transportation technology.

This article proposes a novel 3-D surface temperature prediction model based on the restored pseudoheat flux (RPHF) theory. The method can be used to simulate the temperature difference between the subsurface defect and the sound area. The proposed model shows the potential to investigate the detection limits associated with the defect features, such as depth, radius, diameter-to-depth ratio (D2dR), and excitation features, which is beneficial for the experimental design. Several experiments were conducted on specimens of different materials [glass fiber reinforced plastic (GFRP), CFRP, and rubber] using RPHF thermography to validate the practicality of the model. The comparative analysis is also conducted with other methods. Both experimental and simulation results demonstrated that longer heating is required for deeper defects and the moment of maximum temperature difference tends to appear after the heating has stopped. Probability of detection (PoD) was used as an index to assess the reliability of the methodology and the problems found. The depth of the defect has a greater influence on thermal detection than D2dR.

In this study, we propose a new post processing algorithm, using a stable low-rank decomposed pseudo restored heat flux based on the ensemble variational Bayes tensor factorization (EVBTF-RPHF) algorithm for performing periodic square wave thermographic nondestructive testing (thermographic NDT). Previous studies have shown that both RPHF and EVBTF can separately improve the detectability of thermography by enhancing some defect features. However, both methods are limited by their particularly constraints: RPHF are heavily degraded by noises and missing data due to the assumptions under which the physical models are derived while efficiency of EVBT reduces when the lateral heat diffusion weights out. By embedding RPHF into the stable low-rank decomposition EVBTF, the proposed algorithm allows to improve the detectability of defects in thermographic NDT using a periodic heat flux with low-rank spatial distribution. The study verifies the capacity of the proposed method by theoretical analysis. Then, experiments were conducted on a carbon fiber composite panel with foreign inserts buried up to 5 mm deep. The sampled data are processed by the proposed method. The results are compared with existing methods such as phase-locked RPHF and EVBTF. The experimental results demonstrated that defects with normalized diameter-to-depth ratios as small as 0.9, barely detected with other available techniques, can reliably be detected by EVBTF-RPHF. The signal to noise ratio and the contrast are used as figure of merit to quantitatively compare the capacity of the proposed method with existing methods. However, the computation efficiency of the proposed algorithms needs further improvement.

The future advancement of technology in Internet of Things (IoT) paradigm, Wireless Sensor Networks (WSNs) provide sensing services to connect all the devices. In the upper layer of OSI model designing an energy efficient routing protocol in WSN is a challenge, which can ease the work of Multi-access edge computing (MEC) in IoT applications. The advent of 6G is also playing key role for reliable communication between the sensing elements for IoT applications. These two phenomena are significantly influencing for the progress of next generation Intelligent Transportation System (ITS). Therefore, the proposed work presents a novel method of implementing Distributed Artificial Intelligence (DAI) with neural networks for energy efficient routing as well as a fast response for intra-cluster communication of the nodes to overcome the challenges for ITS. Although there exist several works on the inter-cluster energy-efficient network, our work proposes a new way of implementing the hybrid approach of DAI and Self Organizing Map (SOM). The proposed approach proves to be a better solution in terms of overall energy consumption by the network, along with the computational challenges. Further, the work presents mathematical analysis, simulation results and comparison with the conventional techniques for justification.

To solve the problems of low efficiency and manual dependence of industrial motor winding testing, a joint scanning electromagnetic thermographic (JSET) method and a new quantitative evaluation algorithm are proposed to inspect defects automatically and assess detection capability. We establish a JSET-based defect inspection system including a joint scanning model and induction heating to simulate industrial assembly lines and acquire real-time thermograms. However, the acquired thermograms are misaligned in time and space, which cannot be used for dimension analysis. Therefore, a new 3-D data reconstruction algorithm is proposed to achieve accurate spatial-temporal alignment of the image sequence. Moreover, the parameters (scanning speed and excitation current) of the developed inspection system are optimized through obtaining the maximum inspection quantity. The new quantitative evaluation algorithm can measure the detection capability of different defects types, sizes, and positions by two features of significance and detected area. Experimental results show that the proposed methods can inspect multiple motor winding defects automatically and enhance the inspect efficiency.

In the process of research, development, production, service, and maintenance of silicon photovoltaic (Si-PV) cells and the requirements for detection technology are becoming more and more important. This paper aims to investigate electromagnetic induction (EMI) and image fusion to improve the detection effect of electrothermography (ET) and electroluminescence (EL) of multidefects in Si-PV cells. First, the principles of ET, EL, and other physical processes including EMI, thermal radiation, and luminescence radiation are analyzed in this paper. ET and EL techniques after EMI improvement are used to detect different defects including scratch, broken gridline, surface impurity, hidden crack, and so on. The qualitative results show that EMI can greatly improve the defect detection ability of ET and EL. Then, an image-fusion rule based on L1 norm is proposed to fuse the sparse vector of the ET and EL images. The integration and complementarity of the two wavelength detection data are achieved. Finally, the image-fusion results of sparse representation (SR) algorithm is compared with discrete wavelet transform, curvelet transform, dual-tree complex wavelet transforms, and nonsubsampled contourlet transform. Five objective evaluation indexes including root mean square error, peak signal-to-noise ratio, correlation coefficient, mutual information, and structural similarity index are used to evaluate the fusion results. Overall evaluation results show that the SR algorithm is superior to the other algorithms.

Thermogram reconstruction methods based on one-dimensional models are widely used in data processing for thermography inspections. However, the surface temperature variances caused by thermal diffusion will be compatible with those caused by delamination whose normalized aspect (diameter-to-depth) ratio is close to 1 when the defect itself is very thin, about 0.15 mm in this research. This phenomenon makes the detection capacity of these methods reduced at such defects with small aspect ratios. The paper proposes a new reconstruction method, phase-locked restored pseudo heat flux (RPHF), for thermography inspection using square-wave optical stimulations. The theoretical analysis shows the independence of the method upon the effect of thermal diffusion blur at defect-free areas. Square-wave thermography tests are conducted on a carbon fiber composite panel with artificial delimitations buried up to 4 mm deep. The method is implemented on a private computer to deconvolute the RPHF kernel from the transformed thermogram data. The data are separated into two sets with a one-period phase shift to each other sequentially; a phase-locked substation is applied between the sets. The global signal-to-noise ratios obtained with the proposed method are compared to those obtained with Busse's lock-in phase images and those with thermographic signal reconstruction. The phase-locked RPHF gives the best global signal-to-noise ratios for normalized aspect ratio at 1.1 when sufficient heat is applied. It's concluded that the thermal diffusion effect at defect-free areas should be considered in thermography inspection for defects with a normalized aspect ratio at 1.1.

Impact damage, caused by low-energy impact, is inevitable during the whole life time of carbon fiber reinforced plastic (CFRP) material. However, the barely visible impact damage (BVID) is difficult to be detected by visual methods. Ultrasonic thermography (UT) is an emerging nondestructive testing technique that visualizes damage in thermal images captured by an infrared (IR) camera when the material is stimulated by ultrasound. However, noise and blurry edges around the high-temperature areas may cause confusion and lead to unreliable results in the thermal images of UT test. In this paper, an impact damage inspection method is proposed based on manifold learning for the CFRP material. Low-power ultrasonic excitation is used for this UT. The IR image sequences are processed as datasets in high-dimensional space. These datasets are reduced to lower dimensions by manifold learning to find the intrinsic structure in the two-dimensional manifold. Each dimension of the embedding manifold correlates highly with one degree of freedom underlying the original pixel: steady and random components. The steady component, which reflects the temperature rise caused by damage, is used for VID and BVID detection. The experimental system was set up, and CFRP plate specimens with different impact damage were tested. All the impact damage could be detected and shown in reconstructed static image with little noise. The proposed method using image sequences could provide a visualized, reliable, and effective impact damage inspection and localization means for CFRP material during manufacturing and in service.