Recognition and segmentation of complex texture images based on superpixel algorithm and deep learning

Han, Yuexing; Yang, Shen; Chen, Qiaochuan

doi:10.1016/j.commatsci.2022.111398

Cited by 7 publications

(4 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The basic process of MSIF technology consists of five main steps: raw data acquisition, measurement data preprocessing, data correlation, data decision making and information fusion [19][20]. Target state information is collected by multiple sensors, and these data are temporally and spatially aligned.…”

Section: Figure 1 Multi-sensor Information Fusion Processmentioning

confidence: 99%

Improved Q-Learning-Based Motion Control for Basketball Intelligent Robots Under Multi-Sensor Data Fusion

Lan,

Dong

2024

IEEE Access

View full text Add to dashboard Cite

With the continuous development of science and technology, the application of artificial intelligence in robot technology, especially in the progress of sports robot technology, has received great attention. This development provides new opportunities for the development of intelligent basketball robots. However, achieving high accuracy and stability in motion control remains a challenge in the research field. The motion control technology of basketball intelligent robots is crucial. The research proposes a study on enhancing Q-learning-based action control technology for basketball robots with multi-sensor data fusion to address the problem of inaccuracy and instability. The first step involves examining how multi-sensor data fusion technology can offer more accurate and comprehensive environmental information that can aid the robot in making more precise action decisions. The second area of study involves the development of enhanced algorithms for controlling basketball robots through improved Q-learning techniques. The goal is to increase the accuracy and stability of the robots' actions. The final focus is on creating a model for controlling the basketball robots' actions, which is based on the improved Q-learning algorithm. This model effectively increases the accuracy and stability of the robots' actions. The experiment was conducted on a computer equipped with Intel Core i9-11900K processor and NVIDIA GeForce RTX 3080 graphics card. MATLAB R2023a was used for system parameterization and simulation, while Webots R2023a software was used to simulate robot actions. The sensor adopted Hokuyo URG-04LX-UG01 laser rangefinder, and the robot model was KUKA youBot. At the same time, in order to improve the control performance of basketball intelligent robots, comparisons were made with existing methods such as PID control, trajectory planning, traditional control methods, multi-sensor fusion technology, and visual perception basketball shooting correction technology. The experimental results demonstrated that the enhanced Q-learning algorithm performed exceptionally well, achieving a recognition rate of 99.35%, a response speed of 67.85%, remarkable performance stability of 97.55%, and a low error rate of 7.65%. The improved Qlearning algorithm's efficiency and stability were confirmed in the application of robot control modeling. Compared to existing methods for controlling basketball robots' motion, the proposed method can handle environmental uncertainty more effectively and improve decision-making accuracy. The research has several advantages in robot motion control. It enhances the applicability of basketball intelligent robots in actual competitive sports and provides technical support for robot strategy formulation, rapid adaptation, and efficient execution in actual competitions. It serves as a crucial source for future studies on technology for controlling robot movements and the practical implementation of intelligent basketball robots. INDEX TERMSMulti-sensor data fusion, Q-Learning algorithm, Motion con...

show abstract

Section: Figure 1 Multi-sensor Information Fusion Processmentioning

confidence: 99%

Improved Q-Learning-Based Motion Control for Basketball Intelligent Robots Under Multi-Sensor Data Fusion

Lan,

Dong

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Local image feature information is fused by multiscale small convolution to generate feature map local‐f 11‐13 . The pixels at each position in the feature map local‐f are the fusion enhancement results of multiple local features.…”

Section: Color Analysis Model Of Deep Learning Algorithmmentioning

confidence: 99%

“…Put the characteristic drawing F6 into 3 × 3 and 5 × 5, the two processed results are pooled globally to extract feature points, and then the extracted local feature points are combined with the feature map F6 to complete feature fusion. Local image feature information is fused by multiscale small convolution to generate feature map local-f. [11][12][13] The pixels at each position in the feature map local-f are the fusion enhancement results of multiple local features. The local feature processing channel focuses on the local information of the image, which is beneficial for extracting features with similar distance.…”

Section: Local Feature Processing Channelmentioning

confidence: 99%

Research on color fusion model of historic and cultural blocks in Shanghai based on deep learning algorithm‐take Tianzifang as an example

Wei

Nie

2022

Concurrency and Computation

View full text Add to dashboard Cite

Color fusion is an extremely important part in the reconstruction design of historic and cultural blocks in the process of China's urban historic and cultural blocks transformation. Based on the deep learning algorithm, this paper constructs the color analysis model of cultural blocks, uses the multiscale feature semantic segmentation algorithm analysis technology of dual channel fusion for color research, extracts and analyzes the color of historic and cultural blocks, thus establishes its color fusion model, analyzes the color defects of Tianzifang historic and cultural scenic spot, and proposes to use plants, buildings and pavements to improve the color fusion degree of Shikumen cultural blocks. K E Y W O R D Sarchitectural color protection, deep learning algorithm, historic and cultural blocks, multiscale feature semantic segmentation THE PRESENT SITUATION OF COLOR COMPOSITION OF HISTORIC AND CULTURAL BLOCKS IN SHANGHAI The importance of color in historic and cultural blocksThe emergence of new materials and technologies has impacted the original construction mode and urban texture of historic and cultural blocks.The architectural color of historic and cultural blocks is an important and relatively stable historic product under the natural conditions and cultural traditions in a specific region. It most directly reflects the characteristics of the blocks that are different from the landscape of other regions. It is an indispensable element in the protection of historic and cultural blocks. The blind worship of international architectural style makes people lose the original cultural memory of the city, lose control over the color style and order, and gradually fall into a loss. Although people have realized the importance of color elements in the protection of historic and cultural blocks to a certain extent, due to the lack of standardized and scientific method guidance, the protectors do not know how to scientifically investigate and study the color of block buildings and make feature extraction, and do not know how to reasonably formulate color protection measures. Therefore, to clarify the architectural color is to grasp the keynote of the historic and cultural blocks. The analysis and study of architectural color will help to provide scientific guiding methods and theoretical basis for the protection of historic and cultural blocks. Color elements of historic and cultural blocksAccording to the many factors that affect the landscape color style in the historic and cultural blocks, the vertical elements of the landscape color style are divided into six parts: building body, landscape greening, pavement, public facilities, plaque advertising and traffic facilities. Among them,

show abstract

“…Consequently, effective denoising becomes paramount before undertaking character segmentation and recognition. Efficient character segmentation methods for textured images can be generally categorized into two classes: deep learning-based methods [5] and variational-based methods [6]. Deep learning methods are usually complex and require substantial datasets and high computational capabilities for training and fine-tuning.…”

Section: Introductionmentioning

confidence: 99%

Denoising on Textured Image Using Total Generalized Variation With Overlapping Group Sparsity Based on Fast Split Bregman Method

Zhang,

Yen

2024

IEEE Access

View full text Add to dashboard Cite

The Total Generalized Variation (TGV) model proves effective in removing texture and background patterns in natural images while suppressing the staircase effect introduced by traditional Total Variation (TV) regularization. Nevertheless, TGV falls short in preserving structural features due to its lack of consideration for such structural features. This paper introduces Overlapping Group Sparsity (OGS) regularization into the TGV model with the specific aim of enhancing denoising, particularly in textured images. By leveraging prior knowledge of sparse structures discernible from first-order and second-order gradients, this model surpasses the conventional TGV models in achieving superior denoising and staircase effect elimination. The model proposed employs a fast split Bregman iteration method to address the L1 regularization problem within the complex TGV model combined with OGS. The experimental results comparing various state-of-the-art denoising TV-and TGV-based models highlight a significant improvement in the denoising performance of the proposed model. Specifically, in comparison to the average values of peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) obtained from other state-of-the-art models, the proposed model demonstrated improvements of 3.5% in PSNR and 3.4% in SSIM.INDEX TERMS Total generalized variation, overlapping group sparsity, texture denoising, split Bregman.

show abstract

Recognition and segmentation of complex texture images based on superpixel algorithm and deep learning

Cited by 7 publications

References 42 publications

Improved Q-Learning-Based Motion Control for Basketball Intelligent Robots Under Multi-Sensor Data Fusion

Improved Q-Learning-Based Motion Control for Basketball Intelligent Robots Under Multi-Sensor Data Fusion

Research on color fusion model of historic and cultural blocks in Shanghai based on deep learning algorithm‐take Tianzifang as an example

Denoising on Textured Image Using Total Generalized Variation With Overlapping Group Sparsity Based on Fast Split Bregman Method

Contact Info

Product

Resources

About