Yanyan Li

This paper presents an all-weather drivable area detection method based on deep learning, addressing the challenges of recognizing unstructured roads and achieving clear environmental perception under adverse weather conditions in current autonomous driving systems. The method enhances the Lite-Mobilenetv2 feature extraction module and integrates a pyramid pooling module with an attention mechanism. Moreover, it introduces a defogging preprocessing module suitable for real-time detection, which transforms foggy images into clear ones for accurate drivable area detection. The experiments adopt a transfer learning-based training approach, training an all-road-condition semantic segmentation model on four datasets that include both structured and unstructured roads, with and without fog. This strategy reduces computational load and enhances detection accuracy. Experimental results demonstrate a 3.84% efficiency improvement compared to existing algorithms.

In the manufacturing process of optical lenses, the lack of monitoring and detection of the central processing process leads to low processing efficiency and difficulty in ensuring product consistency. We propose a digital twin system for alignment processing to address this issue. The system adopts a hierarchical architecture based on the digital twin five-dimensional model, aiming to improve the fidelity and interactivity of the virtual model of the centring lathe by combining dimension-driven virtual models with integrated data and physical models of the turning mechanism. We have successfully achieved the semantic and physical fusion of multi-source heterogeneous data during centring processing using information models and OPC UA-based data interaction methods. In addition, we adopted the VMD-GRU method based on feature fusion for real-time monitoring of critical components of the centring lathe. Finally, we validated the feasibility and effectiveness of the digital twin system for the central lathe through development examples. The application of this system is expected to promote the digital and intelligent development of high-precision optical component processing, providing references including references for related manufacturing fields. In summary, we propose a digital monitoring and detection system for the centring process of optical lens manufacturing. The application of this system will help improve product consistency and processing efficiency while reducing risks and costs in the manufacturing process.

In the manufacturing of optical lenses, the lack of digital monitoring and inspection for centring processing not only results in product consistency not being guaranteed but also leads to inefficiency in processing. A centring processing digital twin system is proposed to address the problems of low virtual visualization and insufficient monitoring capability of centring processing of high-precision optical components. The layered architecture of the system is determined based on the five-dimensional model of the digital twin, coupled with a dimension-driven geometric model and a physical model that integrates data and turning mechanisms to improve the fidelity and interactivity of the virtual model of the centring lathe. The fusion of heterogeneous data from multiple sources in the centring machining process at the semantic and physical levels is realized by using an information model and an OPC UA-based data interaction method. The VMD-GRU method based on feature fusion is used to monitor the key components of the centring lathe. Finally, the feasibility and effectiveness of the proposed method are verified by a case study of the development of a digital twin system for centring lathe, which provides a reference for the digitization and intelligence of the machining process of high-precision optical components.

Deep rock formations in extreme environments are characterized by complex working conditions, various structures, high hardness, and high resistance to compression. However, existing coring techniques leave the cores of deep rock formations vulnerable to residual stresses, resulting in poor fidelity during deep coring. This paper develops a rock-breaking model for the structural parameters of drill bits. It proposes that a drill bit’s structural parameters in terms of back-rake and side-rake angles will affect the core’s fidelity performance. In addition, the core’s mechanical specific energy and maximum stress will reflect the fidelity effect. The accuracy of the theoretical model was verified via simulation analysis. The simulation results show that the tool’s average cutting force and Standard deviation of cutting force increase as the drill bit’s back-rake and side-rake angles vary. This leads to increased shear friction on the core, which affects the maximum stress and mechanical specific energy, and, subsequently, the fidelity of the core. The back-rake angles ranged from 15° to 25°, with the optimum back-rake angle of 21° producing a maximum stress and a mechanical specific energy that were 0.69 and 0.85 times higher than the highest point, respectively. The side-rake angles range from 5° to 15°, with the optimum side-rake angle of 10° producing a maximum stress and a mechanical specific energy that were 0.76 and 0.96 times higher than the highest point, respectively. The finite element method error was 1.21%. This work’s main results will help reveal the fidelity mechanisms of the drilling process and contribute to the development of fidelity drill bits for complex surface drilling processes.

This paper proposes a new low‐light image enhancement method, which we call the Local Discrete Mapping Method. The new method limits the processing range to small areas with a high information relevance, which can better coordinate the enhancement quality of each area. First, the discrete mapping relationship of pixels (called discrete mapping points) globally occupying a small part of the critical gray value was extracted and designated to keep the enhancement amplitude of each local area consistent. Then, other free mapping points were adjusted according to the local features to achieve the best visual effect in each local area. In addition, this paper also proposes a span correction function that takes the gray span between local pixels as the adjustment object. The function can preserve the gray difference between freely mapped pixels to the maximum and significantly reduce detailed damage in the local area. Finally, we used 1500 test images and eleven objective evaluation indicators in the public dataset to comprehensively test the seven methods. The experimental results showed that the proposed method has an excellent dark area quality enhancement, brightness detail protection, overall noise suppression, and processing speed. It is significantly better than similar methods in terms of visual quality and quantitative testing.

The density‐based spatial clustering of application with noise (DBSCAN) algorithm has good robustness and is widely employed to cluster vehicle trajectories for vehicle movement pattern recognition. However, the distance or similarity between two trajectories varies from tens to hundreds of thousands, and there is no effective method for determining the values of the hyperparameters eps and MinPts of DBSCAN. In addition, with increasing sizes of trajectory datasets, some trajectory clustering methods that directly analyse points and line segments incur large computational costs and time overhead. To solve these two dilemmas, the authors propose an effective trajectory dimensionality reduction method and a DBSCAN hyperparameter initial value setting method. The trajectory dimensionality reduction algorithm processes trajectories with different lengths into the same dimensionality (the same number of feature points). The reserved points preserve the spatial and temporal information of these trajectories as much as possible. The DBSCAN hyperparameter initial value setting algorithm obtains the effective initial values of eps and MinPts for facilitating subsequent adjustments. Finally, we validate these proposed methods on two trajectory datasets collected from two real‐world scenes, and the experimental results are promising and effective.

A novel night‐time image enhancement approach was proposed in this paper to address the problems of low contrast and poor details of low‐illumination images captured at night. To begin with, the luminance component V was extracted that was irrelevant to the colour information of the image upon converting the image to the HSV space from the RGB space. Then, by converting the luminance component V of the image into the probability space, the image was divided into a small‐probability grey‐scale area and a normal area based on the theory of probability. Moreover, pixels were transferred from the small probability area to the normal area of the image according to the nearest attribution principle that was established. Lastly, the contrast enhancement of the image was realized thanks to lossless mapping functions without losing the number of grey levels of the image. As can be observed from experimental results, the proposed method is superior to the most advanced algorithm in visual quality and quantitative measurement.

We proposed the Retinex-based fast algorithm (RBFA) to achieve low-light image enhancement in this paper, which can restore information that is covered by low illuminance. The proposed algorithm consists of the following parts. Firstly, we convert the low-light image from the RGB (red, green, blue) color space to the HSV (hue, saturation, value) color space and use the linear function to stretch the original gray level dynamic range of the V component. Then, we estimate the illumination image via adaptive gamma correction and use the Retinex model to achieve the brightness enhancement. After that, we further stretch the gray level dynamic range to avoid low image contrast. Finally, we design another mapping function to achieve color saturation correction and convert the enhanced image from the HSV color space to the RGB color space after which we can obtain the clear image. The experimental results show that the enhanced images with the proposed method have better qualitative and quantitative evaluations and lower computational complexity than other state-of-the-art methods.

The width of convolutional neural networks (CNNs) is crucial for improving performance. Many wide CNNs use a convolutional layer to fuse multiscale features or fuse the preceding features to subsequent features. However, these CNNs rarely use blocks, which consist of a series of successive convolutional layers, to fuse multiscale features. In this paper, we propose an approach for improving performance by fusing the low-level features extracted from different blocks. We utilize five different convolutions, including 3×3, 5×5, 7×7,5×3 ∪ 3×5 and 7×3 ∪ 3×7, to generate five low-level features, and we design two fusion strategies: low-level feature fusion (L-Fusion) and high-level feature fusion (H-Fusion). Experimental results show that the L-Fusion is more helpful for improving the performance of CNNs, and the 5×5 convolution is more suitable for multiscale feature fusion. We summarize the conclusion as a strategy that fuses multiscale features in the preceding stage of CNNs. Furthermore, we propose a new architecture to perceive the input of CNNs by using two self-governed blocks based on the strategy. Finally, we modify five off-the-shelf networks, DenseNet-BC (depth = 40), ALL-CNN-C (depth = 9), Darknet 19 (depth = 19), Resnet 18 (depth = 18) and Resnet 50 (depth = 50), by utilizing the proposed architecture to verify the conclusion, and these updated networks provide more competitive results.