Dual stream fusion network for underwater image enhancement of multi-scale turbidity restoration and multi-path color correction.

Underwater images frequently experience color distortion and blurred details due to the absorption and scattering of light, which can hinder underwater visual tasks. To address these challenges, we propose a dual-stream fusion network for enhancing underwater images. Our multi-scale turbidity restoration module (MTRM) adopts a two-stage dehazing process from coarse to fine, while employing the SOS boosting strategy and frequency-based dense connections to further improve the performance of the U-Net. The multi-path color correction module (MCCM) utilizes the multi-path residual block as the basic unit to construct RGB enhancement paths. It selectively establishes inter-color channels through attention-based cross connections, which efficiently harness the distinctive features from various color channels. Additionally, non-local spatial and channel attention provide essential correlation information for the final fusion stage. Qualitative and quantitative evaluations conducted on various underwater datasets have demonstrated the excellent performance of our method.

An Adaptive Multi-Scale Network Based on Depth Information for Crowd Counting.

Crowd counting, as a basic computer vision task, plays an important role in many fields such as video surveillance, accident prediction, public security, and intelligent transportation. At present, crowd counting tasks face various challenges. Firstly, due to the diversity of crowd distribution and increasing population density, there is a phenomenon of large-scale crowd aggregation in public places, sports stadiums, and stations, resulting in very serious occlusion. Secondly, when annotating large-scale datasets, positioning errors can also easily affect training results. In addition, the size of human head targets in dense images is not consistent, making it difficult to identify both near and far targets using only one network simultaneously. The existing crowd counting methods mainly use density plot regression methods. However, this framework does not distinguish the features between distant and near targets and cannot adaptively respond to scale changes. Therefore, the detection performance in areas with sparse population distribution is not good. To solve such problems, we propose an adaptive multi-scale far and near distance network based on the convolutional neural network (CNN) framework for counting dense populations and achieving a good balance between accuracy, inference speed, and performance. However, on the feature level, in order to enable the model to distinguish the differences between near and far features, we use stacked convolution layers to deepen the depth of the network, allocate different receptive fields according to the distance between the target and the camera, and fuse the features between nearby targets to enhance the feature extraction ability of pedestrians under nearby targets. Secondly, depth information is used to distinguish distant and near targets of different scales and the original image is cut into four different patches to perform pixel-level adaptive modeling on the population. In addition, we add density normalized average precision (nAP) indicators to analyze the accuracy of our method in spatial positioning. This paper validates the effectiveness of NF-Net on three challenging benchmarks in Shanghai Tech Part A and B, UCF_ CC_50, and UCF-QNRF datasets. Compared with SOTA, it has more significant performance in various scenarios. In the UCF-QNRF dataset, it is further validated that our method effectively solves the interference of complex backgrounds.

Computer Vision Positioning and Local Obstacle Avoidance Optimization Based on Neural Network Algorithm.

Due to the rapid development of social computerization and smart devices, there is an increasing demand for indoor positioning of mobile robots in the robotics field, so it is very important to realize the autonomous navigation of mobile robots. However, in indoor scenes, due to factors such as dark walls, the global positioning system cannot effectively locate, and the broadband and wired positioning technologies used indoors have problems such as base station laying and delay. Computer vision positioning technology has greatly improved the camera hardware due to its simple equipment and low cost. Compared with other sensor cameras, it is less affected by environmental changes, so visual positioning has received extensive attention. Image matching has become the most critical link in visual positioning. The accuracy, speed, and robustness of image matching directly determine the results of visual positioning, so image matching has become the main topic of this study. In this study, the neural network algorithm is systematically optimized, especially for the robot's local obstacle avoidance, and an obstacle data acquisition method based on VGG16 and fast RCNN is proposed. In order to solve the problem that the semantic image segmentation algorithm based on AlexNet and ResNet is difficult to accurately obtain the information of multiple objects, and an image semantic segmentation algorithm combined with VGG16 is designed to classify the background and road in the image at the pixel level and capture the path boundary line. The collection of robot obstacle path information improves the speed and accuracy of highly automated local obstacle avoidance. This study uses neural network algorithms to systematically optimize computer vision positioning and also studies the accuracy optimization of local obstacle avoidance, aiming to promote its better development.

Dual-flow network with attention for autonomous driving.

We present a dual-flow network for autonomous driving using an attention mechanism. The model works as follows: (i) The perception network extracts red, blue, and green (RGB) images from the video at low speed as input and performs feature extraction of the images; (ii) The motion network obtains grayscale images from the video at high speed as the input and completes the extraction of object motion features; (iii) The perception and motion networks are fused using an attention mechanism at each feature layer to perform the waypoint prediction. The model was trained and tested using the CARLA simulator and enabled autonomous driving in complex urban environments, achieving a success rate of 74%, especially in the case of multiple dynamic objects.

QoE collaborative evaluation method based on fuzzy clustering heuristic algorithm.

At present, to realize or improve the quality of experience (QoE) is a major goal for network media transmission service, and QoE evaluation is the basis for adjusting the transmission control mechanism. Therefore, a kind of QoE collaborative evaluation method based on fuzzy clustering heuristic algorithm is proposed in this paper, which is concentrated on service score calculation at the server side. The server side collects network transmission quality of service (QoS) parameter, node location data, and user expectation value from client feedback information. Then it manages the historical data in database through the "big data" process mode, and predicts user score according to heuristic rules. On this basis, it completes fuzzy clustering analysis, and generates service QoE score and management message, which will be finally fed back to clients. Besides, this paper mainly discussed service evaluation generative rules, heuristic evaluation rules and fuzzy clustering analysis methods, and presents service-based QoE evaluation processes. The simulation experiments have verified the effectiveness of QoE collaborative evaluation method based on fuzzy clustering heuristic rules.

A novel medical image protection scheme using a 3-dimensional chaotic system.

Recently, great concerns have been raised regarding the issue of medical image protection due to the increasing demand for telemedicine services, especially the teleradiology service. To meet this challenge, a novel chaos-based approach is suggested in this paper. To address the security and efficiency problems encountered by many existing permutation-diffusion type image ciphers, the new scheme utilizes a single 3D chaotic system, Chen's chaotic system, for both permutation and diffusion. In the permutation stage, we introduce a novel shuffling mechanism, which shuffles each pixel in the plain image by swapping it with another pixel chosen by two of the three state variables of Chen's chaotic system. The remaining variable is used for quantification of pseudorandom keystream for diffusion. Moreover, the selection of state variables is controlled by plain pixel, which enhances the security against known/chosen-plaintext attack. Thorough experimental tests are carried out and the results indicate that the proposed scheme provides an effective and efficient way for real-time secure medical image transmission over public networks.