Xin-Yu Ouyang

In the past three years, the stereo matching method based on 3D CNNs has achieved surprising results and has received more and more attention. However, most stereo matching approaches aim to improve prediction accuracy by constructing and aggregating cost volumes through extensive 3D convolutions, which not only does not fully utilize the geometric information, but also overlooks the computational speed. Thus, achieving high-accuracy, high-efficiency stereo matching has become challenging. In this paper, we present a rapid and precise stereo matching network named CGW based on 3D CNNs, which simultaneously achieves real-time functioning, considerable accuracy, and a strong generalization capability. The network is divided into two parts. The first part constructs the geometric attention cube through a lightweight feature extraction network and a lightweight 3D regularization network. The second part filters the context features using the geometric attention cube to obtain the context geometric cube, and finally, the disparity is predicted and refined to obtain the final disparity. We adopted MobileNetV3 as an efficient backbone for feature extraction and designed 3D depthwise separable convolutions with residual structures to replace traditional 3D convolutions for constructing the cost volume and performing cost aggregation, aiming to reduce the model size and improve the computational speed. Additionally, we designed the context geometric attention (CGA) module and embedded it into the lightweight 3D regularization network, as well as designed the Warped Disparity Refinement (WDR) network to further improve the disparity prediction accuracy. CGA effectively guides cost aggregation by integrating rich contextual and geometric information, while also providing feedback for feature learning to guide more efficient context feature extraction. WDR constructs a warping cost volume using the obtained initial disparity, combined with image features, the initial disparity map, and reconstruction errors, to optimize the disparity. According to the initial disparity, it searches for the accurate disparity within a refined range. By narrowing the search range, WDR simplifies the task for the network to locate the correct disparity (residual), while simultaneously improving the computational efficiency. Experiments conducted on multiple benchmark datasets showed that, compared to other fast methods, CGW has advantages in both speed and accuracy and exhibits better generalization performance.

In complex field environments, traffic sign detection faces many challenges, such as the effects of light variations, occlusion and sensor resolution, which can lead to a decrease in detection accuracy. To cope with these problems, a multi-information attention fusion traffic sign detection network MIAF-Net is proposed. Firstly, a backbone network with linear transformations was designed to improve the efficiency and accuracy of feature extraction. Secondly, an attention balance feature pyramid network was designed to enhance the correlation between foreground features and surrounding semantics, refine and balance semantic features, and improve the expressive ability of feature maps by fusing and learning multi-scale information. Finally, a detection head with multi-scale information fusion is designed to provide different features for category prediction and boundary regression, increasing the reliability of traffic sign detection and classification. In the experimental part, three traffic sign datasets (TT100K, CCTSDB and DFG) were used to fully evaluate MIAF-Net and compare it with existing state-of-the-art traffic sign detection methods, the results show that MIAF-Net exhibits a very superior performance in the traffic sign detection task. In addition, in the edge device deployment experiments, MIAF-Net demonstrates its real-time performance and low memory access, which shows that the proposed method is not only superior in accuracy, but also has good deployment utility.

Aiming at the tracking control problem of non-affine stochastic nonlinear systems with input saturation constraints, the event trigger control (ETC) based on Radial basic function neural networks (RBFNNs) and prescribed performance control (PPC) is considered. First, the mean value theorem is used to decouple the non-affine terms existing in the system. Second, the design process of PPC is reconstructed for a class of stochastic nonlinear systems, because the existence of stochastic disturbances has not been fully considered in previous literature on PPC, so that the system output is not to violate the set constraint bound by preset function. Finally, unlike the existing event triggering results, a special event triggering strategy is designed, which further takes into account the error between the preset function and the system output and the errors in all states of stochastic nonlinear systems, so it is expected that the amount of communications will be further reduced. Also, the proposed adaptive PPC scheme with event triggering mechanism can guarantee that all closed-loop signals are uniformly ultimately bounded (UUB) in probability within the appropriate compact sets. Finally, the effectiveness of the proposed method is verified by a practical example.