Vision-centric autonomous driving systems require diverse data for robust training and evaluation, which can be augmented by manipulating object positions and appearances within existing scene captures. While recent a...
A crop disease detection method suitable for modern agricultural greenhouses is proposed in this paper. Based on the principle of machine vision detection technology and LabVIEW software program, the disease recogniti...
A crop disease detection method suitable for modern agricultural greenhouses is proposed in this paper. Based on the principle of machine vision detection technology and LabVIEW software program, the disease recognition and detection on the surface of potted crops can be quickly achieved. The system first preprocesses the image, and through processes such as system calibration and image correction, grayscale histogram analysis, threshold segmentation, particle analysis and processing, feature analysis, and pattern matching, it achieves rapid recognition of crop diseases. When performing threshold segmentation on images, this system introduces a background correction algorithm to preprocess the images. The image recognition method for crop leaf diseases studied in this article makes monitoring and formulating pest control measures more effective.
作者：
Wang, HaoyuanFu, ZhongzhengLi, AndongYi, LujieSun, YuxiaoChen, XinxingWu, HaoHuang, Jian
Hubei Key Laboratory of Brain-inspired Intelligent Systems Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Hubei Wuhan430074 China
State Key Laboratory of Intelligent Manufacturing Equipment and Technology School of Mechanical Science and Engineering Hubei Wuhan430074 China
Current research on grasping state analysis in soft manipulators is limited and lacks broad applicability. In this article, we introduce a novel method that leverages multimodal data from flexible sensors and Inertial...
详细信息
A trend in structured light 3D measurement systems is miniaturization. To address this trend, this paper presents a high-speed structured light vision imaging and storage system based on Xilinx Zynq-7000 SoC. The syst...
详细信息
Particle filters (PFs) are a set of Bayesian approaches to estimate the posterior densities of the state of the systems from given observations using a set of weighted samples. Due to their superiority in dealing with...
Particle filters (PFs) are a set of Bayesian approaches to estimate the posterior densities of the state of the systems from given observations using a set of weighted samples. Due to their superiority in dealing with non-linear and non-Gaussian systems, PFs are widely used in real-time applications such as localization and tracking. However, they are computationally demanding for real-time applications since large number of particles are used. Besides, the filtering performance doesn’t always increase as the number of particles increases, and redundant particles increase the computational complexity. Towards this problem, this paper proposes an adaptive particle filter method which adapts the size of particle set dynamically based on its effective sample size (ESS) under current observations. This method enables effective number of particles used for real-time target tracking while maintaining the tracking quality. Experimental results demonstrate that the proposed method achieves significant particle reduction, which is up to 33% under the same tracking quality compared to the standard PFs, confirming the effectiveness of our proposed method in terms of both tracking performance and speed.
Automatic and accurate segmentation of the left atrium (LA) is a prerequisite for quantifying the LA, and effective LA segmentation is helpful for the clinical diagnosis of patients with atrial fibrillation. The exist...
详细信息
Automatic and accurate segmentation of the left atrium (LA) is a prerequisite for quantifying the LA, and effective LA segmentation is helpful for the clinical diagnosis of patients with atrial fibrillation. The existing LA segmentation methods are prone to a lack of structural prediction of the pulmonary vein and mitral valve, and the structure of the LA is prone to over-and under-segmentation. To solve these problems, we propose a comprehensive information integration network (CII-Net) to segment the LA from late gadolinium-enhanced (LGE) cardiac magnetic resonance (CMR) images. The CII-Net integrates multiscale information from the input stage, high-level semantic information from the bottleneck stage, interaction information from the encoding and decoding stages, and information from the output stage, which can more completely capture the comprehensive image characteristics. We conducted experiments on 154 LGE CMR cases with atrial fibrillation proposed by the organizer of the 2018 atrial segmentation challenge. Compared with state-of-the-art methods, the proposed CII-Net obtains an average DICE score of 91.9%, average Jaccard score of 85.1%, average Hausdorff distance of 5.924 mm, and average symmetric surface distance of 0.993 mm without post-processing, demonstrating the potential for clinical application.
The central idea of contrastive learning is to discriminate between different instances and force different views from the same instance to share the same representation. To avoid trivial solutions, augmentation plays...
详细信息
In view of the current situation that the traditional prediction of leaf disease information during the growth of citrus has low accuracy and complicated prediction methods, this paper proposes a citrus leaf disease a...
详细信息
In this study, a hybrid online brain-computer interface (BCI) system was established to address the issue of low recognition accuracy and limited imagery tasks in motor imagery (MI) based on BCI systems. This hybrid s...
In this study, a hybrid online brain-computer interface (BCI) system was established to address the issue of low recognition accuracy and limited imagery tasks in motor imagery (MI) based on BCI systems. This hybrid system utilized both MI and motion-onset visual evoked potential (mVEP). After simultaneously acquiring MI and mVEP potentials and electroencephalogram (EEG) signal preprocessing, the filter bank common spatial pattern (FBCSP) algorithm was used to extract MI and mVEP features. These features were then employed to construct training set feature vectors. The training set feature vectors of MI and mVEP signals were classified using the backpropagation (BP) neural network and support vector machine (SVM) algorithms, respectively. Based on the established classification model, single test data were recognized online, and the results were fed back to subjects in real-time via block moving routes. The presented system can realize five different block moving paths, with faster moving effects achieved when the system induces MI and mVEP potential simultaneously along the lower left or lower right diagonal path. The experimental results showed that the average classification accuracies of the lower left and lower right diagonal moving effects are up to 84.53% and 84.60%, respectively. These results indicate that this study effectively controls the fast-moving blocks in the online BCI system of MI-mVEP, providing a theoretical basis and experimental support for the research of a hybrid online BCI system based on MI-mVEP. In summary, the proposed system has demonstrated improved performance through the combination of simultaneous MI and mVEP potential utilization and efficient classification algorithms, with promising future applications in BCI research.
Flatness is a significant indicator to evaluate the quality of medium-thick steel plates, which affects subsequent processing and transportation. The existing detection is mainly based on single - line laser vision, p...
详细信息
暂无评论