Nowadays, autonomous vehicle is an active research area, especially after the emergence of machine vision tasks with deep learning. In such a visual navigation system for vehicle, controller captures images and predicts information so that can safely navigate. this paper, we first introduced small medium-sized obstacles were intentionally or unintentionally left on road, which pose hazards both human driving situations. Then, discuss Markov random field (MRF) model by fusing three potentials...

In substation working environments, analog dial gauges are widely used for equipment monitoring. Accurate reading of values is crucial real-time understanding operational status and enhancing the intelligence operation maintenance. However, existing recognition algorithms face significant errors in complex scenarios struggle to adapt dials with different measurement ranges. To address these issues, this paper proposes an automatic method consisting two stages: segmentation recognition....

In practical applications, the clarity of analog dial images is often compromised due to factors such as lighting conditions, leading low precision and poor segmentation scales pointers. This results in outcomes that fail meet real-time requirements substation inspection systems. To address these challenges, we propose an improved U-Net algorithm. The key innovation our approach insertion a layer-hopping connection module between Encoder Decoder capture feature information across multiple...

Existing substation equipment image data augmentation models face challenges such as high dataset size requirements, difficult training processes, and insufficient condition control. This paper proposes a transformer method based on Stable Diffusion model. The proposed incorporates the Low-Rank Adaptation (LoRA) concept to fine-tune pre-trained model weights, significantly reducing requirements while effectively integrating essential features of data. To minimize interference from complex...

Image data is as common textual in this digital world. There an urgent demand of image management tools efficient those text search engines. Decades research on retrieval has found there a significant gap between the existing content based and semantic interpretation images by human. As result, recent shifted to retrieval. Many models have been proposed, however, these methods are still alienated from widely accepted method. In paper, we propose unite model with using novel region inverted...

There has been a growing interest in implementing image search engine at the semantic level. However, most existing practical systems including popular commercial engines like Google and Yahoo! are either text-based or simple hybrid of

This paper proposes an efficient lane markings detection and tracking method, which utilizes line segments as feature information combined with vanishing point constraints. The method can be highlighted in four items follows. Firstly, the region of interest (ROI) road image is determined, then edge extracted by Canny detector. Secondly, scanned to calculate orientation edge-linking pixels we filter out noise edges have abnormal orientation. Then, detected Progressive Probabilistic Hough...

The lane lines’ length, width, and direction are very regular, serialized, structurally associated, which not easily affected by the environment. To enhance detection in a complicated environment, an approach combines visual information with spatial distribution. Firstly, grid density of target algorithm YOLOv3 (you only look once V3) is improved from S×S to S×2S, aiming at particular points bird’s-eye view where lines had different densities horizontal vertical directions. obtained (S×2S)...

Recently, semantic image retrieval has attracted large amount of interest due to the rapid growth digital storage. However, existing approaches have severe limitations. In this paper, a new approach using intermediate features and multistep search been proposed. Instead looking for human level semantics which is too challenging at stage, research looks heuristic information can describe content objectively. Different from conventional approaches, are used as filters eliminate irrelevant...

With the rapid development of digital imaging technology, image annotation is an important and challenging task in retrieval. At present, many machine learning methods have been applied to solve problem automatic (AIA). However, there exists enormous semantic expressive gap between low-level features high-level concepts. Due problem, performance existing not satisfactory, needs be further improved. This paper proposes framework via a novel decision tree-based Bayesian (DTB) algorithm. It...

The residual neural network is prone to two problems when it used in the process of face recognition: first "overfitting", and other slow or non-convergence problem loss function later stage training. In this paper, order solve paper increases number training samples by adding Gaussian noise salt pepper original image achieve purpose enhancing data, then we added "dropout" network, which can improve generalization ability network. addition, have improved optimization algorithm After...

Hands emerge quite often in egocentric video, and their structure disposition reveal a lot as to what individuals have been engaged showing interest to. Earlier researches hand detection, on the other hand, depended significant effect that only operate limited situations, such lab settings or with minimal social interaction. This work investigates recently developed YOLO architectural for object recognition which offers good balance of precision performance. We performed an experiment...

For efficiency problem of virtual human hair modeling and simulation process in technology, an improved collision detection algorithm for study is proposed. The triangle fast interpolation method used Aixe Align Bounding Boxes(AABB) spheres are to deal with the hair. Finally, interaction handled hybrid Eulerian/Lagrangian approach. Collision only detect state between controlled head model. interpolated on basis after treatment, so that no performed. experiment indicates can work effectively...

Feature-based intelligent design and manufacturing systems in the Internet environment are an evolution of traditional geometric solid modeling systems. This paper presents some novel algorithms including a new face-base representation, composite feature retrieval technology, efficient communication mechanism, to construct interactive framework for retrieval. The proposed system consists modeler developed on Wolfram Research Mathematica, Java 3D enabled GUI (graphical user interface), DB...