Lei Huang

Vessel detection and tracking is of utmost importance to river traffic. Efficient detection and tracking technology offer an effective solution to address challenges related to river traffic safety and congestion. Traditional image-based object detection and tracking algorithms encounter issues such as target ID switching, difficulties in feature extraction, reduced robustness due to occlusion, target overlap, and changes in brightness and contrast. To detect and track vessels more accurately, a vessel detection and tracking algorithm based on the LiDAR point cloud was proposed. For vessel detection, statistical filtering algorithms were integrated into the Euclidean clustering algorithm to mitigate the effect of ripples on vessel detection. Our detection accuracy of vessels improved by 3.3% to 8.3% compared to three conventional algorithms. For vessel tracking, L-shape fitting of detected vessels can improve the efficiency of tracking, and a simple and efficient tracking algorithm is presented. By comparing three traditional tracking algorithms, an improvement in multiple object tracking accuracy (MOTA) and a reduction in ID switch times and number of missed detections were achieved. The results demonstrate that LiDAR point cloud-based vessel detection can significantly enhance the accuracy of vessel detection and tracking.

To acquire an accurate location on the occasions, such as in an indoor, tunnel, and valley, where satellite navigation signals fail. The paper designs a pedestrian navigation system by using the zero velocity update procedure technology (ZUPT) and Kalman filter to reduce the location error. The measurement noise characteristic (mean and variance) of the micro electro mechanical systems gyros is unknown and time variant, but in traditional studies, it is usually thought and calculated as a constant. So the optimality of the error estimation of the Kalman filter cannot be reached. To address this question, this paper proposes the improved Sage–Husa Adaptive Kalman Filter (SHAKF) based on the index fading memory factor to realise the state estimation of the Kalman filter and navigation error correction. The advantage of improved SHAKF is it can accurately estimate the state vector when the measurement noise is unknown and time variant. To verify the validity of novel navigation methods, walking experiments under outdoor environments and indoor environments are carried out. The results of actual walking experiments demonstrate that the proposed method can effectively reduce the pedestrian location error compared with the traditional ZUPT method. The mean location error is reduced by more than 10%, and the variance of the location error is reduced by more than 5%.

In order to improve the accuracy of visual SLAM algorithms in a dynamic scene, instance segmentation is widely used to eliminate dynamic feature points. However, the existing segmentation technology has low accuracy, especially for the contour of the object, and the amount of calculation of instance segmentation is large, limiting the speed of visual SLAM based on instance segmentation. Therefore, this paper proposes a contour optimization hybrid dilated convolutional neural network (CO-HDC) algorithm, which can perform a lightweight calculation on the basis of improving the accuracy of contour segmentation. Firstly, a hybrid dilated convolutional neural network (HDC) is used to increase the receptive field, which is defined as the size of the region in the input that produces the feature. Secondly, the contour quality evaluation (CQE) algorithm is proposed to enhance the contour, retaining the highest quality contour and solving the problem of distinguishing dynamic feature points from static feature points at the contour. Finally, in order to match the mapping speed of visual SLAM, the Beetle Antennae Search Douglas–Peucker (BAS-DP) algorithm is proposed to lighten the contour extraction. The experimental results have demonstrated that the proposed visual SLAM based on the CO-HDC algorithm performs well in the field of pose estimation and map construction on the TUM dataset. Compared with ORB-SLAM2, the Root Mean Squared Error (Rmse) of the proposed method in absolute trajectory error is about 30 times smaller and is only 0.02 m.

In recent years, the problem of image complementation has achieved better complementation effect under the condition of applying deep learning technology, and even the human eye is difficult to distinguish. Therefore, the technology has also become a hot research topic on image complementation. At the same time, how to modify the generative adversarial network to better adapt to the problem of image complementation to construct a more efficient generation model has been paid more and more attention. Face complementation, as a branch of the problem of picture complementation, is a common face image editing technique, which can also be used to edit face properties. The resulting face image can be as accurate as the original face image, or it can be consistent with the content of the unobstructed face image so that the completed image appears to have a real visual feel. Typically, the Generative Adversarial Network can generate missing parts of the face, but the results generated are often incongruous with the entire face, the texture is incoherent, the context is not integrated, and there are obvious signs of repair at the connection between the missing area and the non-missing area. This paper proposes the mode of combining Generative Adversarial Network and Texture Network. Generative Adversarial Network by training a large number of samples, generating an image of the face mask area, and then Texture Network smoothing the image, adding textures, making the resulting face more realistic. Experimental results from the CelebA dataset show that our model can handle a large number of missing pixels and produce realistic facial completion results.

In order to solve the problem of low face recognition accuracy of traditional algorithms and excessive long training time in deep learning methods, a novel lightweight and short training time multiface recognition method is proposed in this paper. Firstly, an extraction model of facial feature vectors is established based on local binary mode and principal component analysis. Secondly, the beetle antennae search algorithm (BAS) is optimized using adaptive factors, and the ABAS algorithm is proposed. This paper uses the ABAS algorithm to optimize the initial threshold of the neural network and proposes the ABASNet method. Thirdly, ABASNet method is combined with the face features extracted by the face feature vector extraction model to be used for multi-face classification tasks. Finally, H-softmax (hierarchical softmax) is used to replace softmax in neural networks. By reducing the amount of calculation in the process of multi-category face classification, the training time of the ABASNet network is reduced. Through ablation experiments and comparative experiments with various methods such as IKDA + PNN algorithm and PCANet algorithm, to verify the accuracy, robustness and short training time of the ABASNet method. Among them, the maximum accuracy of the method in the ORL face dataset, ExtYaleB face dataset and FERET face dataset is 99.35%, 99.54% and 99.18% respectively. In addition, actual test results indicate that the training time and recognition time of the method in this paper are 21 seconds and 0.06 seconds respectively, which has demonstrated the lightweight and real-time performance of the proposed method. At the same time, more test results show that the proposed ABASNet method can also be implemented in embedded devices and other classification tasks.

Vehicle detection in intelligent transportation systems (ITS) is a very important and challenging task in traffic monitoring. The difficulty of this task is to accurately locate and classify relatively small vehicles in complex scenes. To solve these problems, this paper proposes a modified one-stage detector based on background prediction and group normalization to realize real-time and accurate detection of traffic vehicles. The one-stage detector firstly adds a module to adjust the width and height of anchors and predict the target background, which avoids the problem of the target vehicle missing detection or wrong detection due to the influence of the complicated environments. Then, group normalization and the loss function based on weight attenuation can improve the one-stage detector performance in the training process. The experimental results on traffic monitoring datasets indicate that the improved one-stage detector is superior to the other neural network models in terms of precision at 95.78%.

With the advance of gyro technology, modern gyros have two output types: angular rate or integrated angular rate. However, the conventional sculling velocity algorithms usually still adopt integrated angular rate/specific-force increments as algorithm inputs. So the engineer must convert the angular rate into integrated angular rate by digital integration to use them. This step will produce non-negligible computational error. To solve this issue, we proposed two types of novel optimal sculling algorithms using angular rate input. The advantage of the novel algorithms is that they can directly calculate out the carrier velocity without converting the angular rate of gyro output into integrated angular rate. Hence, they have a higher accuracy than the conventional sculling algorithms. The results of digital simulations also demonstrate this conclusion.

Sculling motion is a standard input to evaluate the performance of the velocity algorithm in a highly dynamic environment. Conventional sculling algorithms usually adopt incremental angle/specific force increments or angular rate/specific force as algorithm inputs. However modern inertial sensors have different output types now, which do not correspond to the inputs of those traditional algorithms. For example, some inertial sensors have the integrated angular rate (incremental angle)/specific force outputs or angular rate/specific force increments outputs. Hence the conventional sculling algorithms cannot be easily applied to these situations. A novel sculling algorithm using incremental angle/specific force inputs or angular rate/specific force increments inputs is developed in this paper. The advantage of the novel algorithm is that it can calculate the carrier velocity directly without converting the dimension of inertial sensor outputs values. Theoretical analysis, digital simulations, and a trial study are carried out to verify our algorithm. The results demonstrate that for corresponding types of strapdown inertial navigation systems (SINS) the novel sculling algorithm exhibits better performance than the conventional sculling velocity algorithms.

In the micro-strapdown inertial navigation systems of Internet of Things, modeling and filtering of gyro random noise are a useful approach to reducing sensor error and enhancing navigation accuracy. Time series is a popular choice for the gyro random noise modeling process. This paper contains two parts to research the modeling methods of gyro random noise. In the first part, for the problem that conventional autoregressive (AR) modeling methods for fiber optic gyro random noise require mass data due to the slow convergence speed of parameter estimation, furthermore, these methods do not perform well if there are too few samples (e.g., less than 3000), and a novel AR modeling method based on the adaptive Kalman filter is proposed. The proposed method can work with smaller sample sizes due to its rapid parameter estimation convergence speed; its algorithm also yields very precise results. These advantages make it well-suited to applications that require a fast AR modeling method for gyro random noise. In the second part, an experiment reveals that the random noise characteristic of some micro-electromechanical systems gyros is adapted to the moving average (MA) model. However, from the available literature, research on a gyro random noise modeling method based on an MA model is very rare. Therefore, we discuss and analyze two MA modeling methods for gyro random noise and give the comparisons between their modeling results.