Real-Time Target Detection Method Based on Lightweight Convolutional Neural Network

Frontiers in Bioengineering and Biotechnology, Journal Year: 2022, Volume and Issue: 10

Published: Aug. 16, 2022

The continuous development of deep learning improves target detection technology day by day. current research focuses on improving the accuracy technology, resulting in model being too large. number parameters and speed are very important for practical application embedded systems. This article proposed a real-time method based lightweight convolutional neural network to reduce improve speed. In this article, depthwise separable residual module is constructed combining convolution non-bottleneck-free module, structure used replace VGG backbone SSD feature extraction parameter quantity At same time, kernels 1 × 3 standard adding 1, respectively, obtain multiple graphs corresponding SSD, established integrating information graphs. self-built dataset complex scenes comparative experiments; experimental results verify effectiveness superiority method. tested video performance model, deployed Android platform scalability model.

Language: Английский

---

Path Planning Optimization of Intelligent Vehicle Based on Improved Genetic and Ant Colony Hybrid Algorithm

Frontiers in Bioengineering and Biotechnology, Journal Year: 2022, Volume and Issue: 10

Published: July 1, 2022

Intelligent vehicles were widely used in logistics handling, agriculture, medical service, industrial production, and other industries, but they often not smooth enough planning the path, number of turns was large, resulting high energy consumption. Aiming at unsmooth path problem four-wheel intelligent vehicle algorithm, this article proposed an improved genetic ant colony hybrid physical model established. This first optimization algorithm about heuristic function with adaptive change evaporation factor. Then, it on fitness function, adjustment crossover factor, mutation Last, addition a deletion operator, adoption elite retention strategy, suboptimal solutions obtained from to obtain optimized new populations. The simulation environment for is windows 10, processor Intel Core i5-5257U, running memory 4GB, compilation MATLAB2018b, samples 50, maximum iterations 100, initial population size 200, 50. Simulation experiments show that effective. Compared traditional reduced by 46% average 75% simple grid. 47% 21% complex works better reduce maps.

Language: Английский

---

Grasping Pose Detection for Loose Stacked Object Based on Convolutional Neural Network With Multiple Self-Powered Sensors Information

IEEE Sensors Journal, Journal Year: 2022, Volume and Issue: 23(18), P. 20619 - 20632

Published: Aug. 5, 2022

There are a variety of objects, random postures and multiple objects stacked in disorganized manner unstructured home applications, which leads to the object grasping posture estimation planning based on machine vision become very complicated. This paper proposes method cluttering pose detection convolutional neural network with self-powered sensors information. Firstly, search strategy for candidate poses 3D point cloud is proposed, single-channel image dataset representing this established by using Bigbird dataset. Secondly, ResNet constructed rank filter single channel captured images bit pose. It also compared three mainstream classification networks, Inception V2, VGG-A LetNet, perception analysis function execution developed under ROS. The effective manipulator scene scattered piles realized results position combined information sensors, other networks. In environment experiment show that superior average success rate ResNet, InceptionV2, VGGA LetNet networks 90.67%, 82.67%, 86.67% 87.33% respectively, verifies effectiveness superiority deep learn-based model proposed paper.

Language: Английский

---

Improved Multi-Stream Convolutional Block Attention Module for sEMG-Based Gesture Recognition

Frontiers in Bioengineering and Biotechnology, Journal Year: 2022, Volume and Issue: 10

Published: June 7, 2022

As a key technology for the non-invasive human-machine interface that has received much attention in industry and academia, surface EMG (sEMG) signals display great potential advantages field of collaboration. Currently, gesture recognition based on sEMG suffers from inadequate feature extraction, difficulty distinguishing similar gestures, low accuracy multi-gesture recognition. To solve these problems new network called Multi-stream Convolutional Block Attention Module-Gate Recurrent Unit (MCBAM-GRU) is proposed, which signals. The multi-stream formed by embedding GRU module CBAM. Fusing ACC further improves action experimental results show proposed method obtains excellent performance dataset collected this paper with accuracies 94.1%, achieving advanced 89.7% Ninapro DB1 dataset. system high classifying 52 kinds different delay less than 300 ms, showing terms real-time human-computer interaction flexibility manipulator control.

Language: Английский

---

An Improved SSD-Like Deep Network-Based Object Detection Method for Indoor Scenes

IEEE Transactions on Instrumentation and Measurement, Journal Year: 2023, Volume and Issue: 72, P. 1 - 15

Published: Jan. 1, 2023

The indoor scene object detection technology is of important research significance, which one the popular topics in field understanding for robots. In recent years, solutions based on deep learning have achieved good results detection. However, there are still some problems to be further studied methods, such as lighting problem and occlusion caused by complexity environment. Aiming at these problems, an improved method neural networks proposed this article, uses a framework similar single-shot multibox detector (SSD). method, ResNet50 network used enhance transmission information, feature expression capability extraction improved. At same time, multiscale contextual information (MCIE) module extract scene, so improve effect. addition, dual-threshold non-maximum suppression (DT-NMS) algorithm alleviate scenes. Finally, public dataset SUN2012 screened special application detection, tested dataset. experimental show that mean average precision (mAP) can reach 54.10%, higher than those state-of-the-art methods.

Language: Английский

---

UAV-based Internet of Vehicles: A systematic literature review

Intelligent Systems with Applications, Journal Year: 2023, Volume and Issue: 18, P. 200226 - 200226

Published: April 19, 2023

The Internet of Vehicles (IoV) is a network Internet-enabled intelligent vehicles outfitted with sensors, customized software, and communication technologies, which result the development vehicular ad hoc networks (VANETs). Unmanned aerial (UAVs) have recently sparked much interest in both VANETs IoV ecosystems. UAVs-enabled techniques can be used to facilitate various applications. This paper investigates UAV-based approaches using systematic literature review (SLR) methodology. We 41 research papers on UAV-enabled published from 2018 2022. Our taxonomy reveals that publications are divided into two main categories, including requirements, processing requirements. results our investigation evaluation factors show delay factor has highest frequency 33%, whereas energy consumption 41%. In this paper, we highlight most significantly utilized subjects ecosystem investigate hot keywords, protocols, implementation strategies, performance metrics. Finally, discuss outstanding issues future challenges domain.

Language: Английский

---

An improved target detection method based on YOLOv5 in natural orchard environments

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 219, P. 108780 - 108780

Published: Feb. 29, 2024

Language: Английский

---

Detection method of wheat spike improved YOLOv5s based on the attention mechanism

Frontiers in Plant Science, Journal Year: 2022, Volume and Issue: 13

Published: Sept. 28, 2022

In wheat breeding, spike number is a key indicator for evaluating yield, and the timely accurate acquisition of great practical significance yield prediction. actual production; method using an artificial field survey to count spikes time-consuming labor-intensive. Therefore, this paper proposes based on YOLOv5s with improved attention mechanism, which can accurately detect small-scale better solve problems occlusion cross-overlapping spikes. This introduces efficient channel module (ECA) in C3 backbone structure network model; at same time, global mechanism (GAM) inserted between neck head structure; be more Effectively extract feature information suppress useless information. The result shows that accuracy model reached 71.61% task number, was 4.95% higher than standard had counting accuracy. YOLOv5m have similar parameters, while RMSE MEA are reduced by 7.62 6.47, respectively, performance YOLOv5l. improves its applicability complex environments provides technical reference automatic identification numbers estimation. Labeled images, source code, trained models available at: https://github.com/228384274/improved-yolov5.

Language: Английский

---

Biomimetic Vision for Zoom Object Detection Based on Improved Vertical Grid Number YOLO Algorithm

Frontiers in Bioengineering and Biotechnology, Journal Year: 2022, Volume and Issue: 10

Published: May 20, 2022

With the development of bionic computer vision for images processing, researchers have easily obtained high-resolution zoom sensing images. The drones equipped with high-definition cameras has greatly increased sample size and image segmentation target detection are important links during process information. As biomimetic remote usually prone to blur distortion in imaging, transmission processing stages, this paper improves vertical grid number YOLO algorithm. Firstly, light shade a were abstracted, grey-level cooccurrence matrix extracted feature parameters quantitatively describe texture characteristics image. Simple Linear Iterative Clustering (SLIC) superpixel method was used achieve light/dark scenes, saliency area obtained. Secondly, model segmenting dark scenes established made dataset meet recognition standard. Due refraction passing through lens other factors, difference contour boundary value between pixel background would make it difficult detect target, pixels main part separated be sharper edge detection. Thirdly, algorithm an improved proposed real time on processed array. adjusted aspect ratio modified grids network structure by using 20 convolutional layers five maximum aggregation layers, which more accurately adapted "short coarse" identified object information density. Finally, comparison mainstream algorithms different environments, test results aid showed that high spatial resolution images, higher accuracy than had real-time performance accuracy.

Language: Английский

---

Tiny object detection model based on competitive multi-layer neural network (TOD-CMLNN)

Intelligent Systems with Applications, Journal Year: 2023, Volume and Issue: 18, P. 200217 - 200217

Published: March 23, 2023

Tiny Object Detection (TOD) is a fundamental and difficult task in computer vision. Current state-of-the-art detectors like RCNN, Fast Faster SSD, YOLO can't find small objects using single-stage or multi-stage methods. With the exponential growth of deep learning, several researchers have drawn attention to advances tiny object detection approaches. This study proposes TOD-CMLNN (Tiny Competitive Multi-Layer Neural Network) architecture with three sub components first competitive multi-layer network, second TOD auxiliary third multi-level continue features aggregation for accurately detecting objects. learning basis proposed architecture. Comparison existing SSD shows significant improvement results. receives 72.46 % accuracy terms mAP which impressive as compared detectors.

Language: Английский

---