Deep Learning Based Vehicle Detection and Counting System for Intelligent Transportation

Computer Systems Science and Engineering, Journal Year: 2024, Volume and Issue: 48(1), P. 115 - 130

Published: Jan. 1, 2024

Traffic monitoring through remote sensing images (RSI) is considered an important research area in Intelligent Transportation Systems (ITSs).Vehicle counting systems must be simple enough to implemented realtime.With the fast expansion of road traffic, real-time vehicle becomes essential constructing ITS.Compared with conventional technologies, sensing-related technique for exhibits greater significance and considerable advantages its flexibility, low cost, high efficiency.But several techniques need help balancing complexity accuracy technique.Therefore, this article presents a deep learning-based detection system ITS (DLVDCS-ITS) images.The presented DLVDCS-ITS intends detect vehicles, count classify them into different classes.At initial level, method applies improved RefineDet detection.Next, Gaussian Mixture Model (GMM) employed process.Finally, sooty tern optimization (STO) convolutional autoencoder (DCAE) model used classification.A brief experimental analysis was made demonstrate enhanced performance technique.The comparative displays method's supremacy over current state-of-the-art approaches.

Language: Английский

---

One to All: Towards a Unified Model for Counting Cereal Crop Heads Based on Few-Shot Learning

Plant Phenomics, Journal Year: 2024, Volume and Issue: 6

Published: Jan. 1, 2024

Accurate counting of cereals crops, e.g., maize, rice, sorghum, and wheat, is crucial for estimating grain production ensuring food security. However, existing methods cereal crops focus predominantly on building models specific crop head; thus, they lack generalizability to different varieties. This paper presents Counting Heads Cereal Crops Net (CHCNet), which a unified model designed multiple heads by few-shot learning, effectively reduces labeling costs. Specifically, refined vision encoder developed enhance feature embedding, where foundation model, namely, the segment anything (SAM), employed emphasize marked while mitigating complex background effects. Furthermore, multiscale interaction module proposed integrating similarity metric facilitate automatic learning crop-specific features across varying scales, enhances ability describe various sizes shapes. The CHCNet adopts 2-stage training procedure. initial stage focuses latent mining capture common representations crops. In subsequent stage, inference performed without additional training, extracting domain-specific target from selected exemplars accomplish task. extensive experiments 6 diverse datasets captured ground cameras drones, substantially outperformed state-of-the-art in terms cross-crop generalization ability, achieving mean absolute errors (MAEs) 9.96 9.38 13.94 7.94 15.62 mixed A user-friendly interactive demo available at http://cerealcropnet.com/, researchers are invited personally evaluate CHCNet. source code implementing https://github.com/Small-flyguy/CHCNet.

Language: Английский

---

VrsNet - density map prediction network for individual tree detection and counting from UAV images

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2024, Volume and Issue: 131, P. 103923 - 103923

Published: June 2, 2024

Individual tree detection and counting in unmanned aerial vehicle (UAV) imagery constitute a vital practical research field. Vegetation remote sensing captures large-scale trees characterized by complex textures, significant growth variations, high species similarity within the vegetation, which presents challenges for annotation detection. Existing methods based on bounding boxes have struggled to convey semantics information about crowns. This paper proposes novel deep learning network called VrsNet density map information. The proposed work pioneers segmentation application utilizing semantic of Gaussian contour. Besides, we sample create UAV vegetation dataset TreeFsc experiments. In quantitative comparison across multiple datasets, method demonstrates performance, with 3.45 increase MAE 4.75 RMSE. Experiments demonstrate superior cross-region, cross-scale, cross-species target capabilities approach compared existing object methods. Our code are available at: https://github.com/luotiger123/VrsNet/tree/main/VrsNet.

Language: Английский

---

An Improved YOLOv8-Based Lightweight Attention Mechanism for Cross-Scale Feature Fusion

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(6), P. 1044 - 1044

Published: March 16, 2025

This paper addresses the challenge of small object detection in remote sensing image recognition by proposing an improved YOLOv8-based lightweight attention cross-scale feature fusion model named LACF-YOLO. Prior to backbone network outputting maps, this introduces a module, Triplet Attention, and replaces Concatenation with Fusion (C2f) more convenient higher-performing dilated inverted convolution layer acquire richer contextual information during extraction phase. Additionally, it employs convolutional blocks composed partial pointwise as main body integrate from different levels. The also utilizes faster-converging Focal EIOU loss function enhance accuracy efficiency. Experimental results on DOTA VisDrone2019 datasets demonstrate effectiveness model. Compared original YOLOv8 model, LACF-YOLO achieves 2.9% increase mAP 4.6% mAPS dataset 3.5% 3.8% dataset, 34.9% reduction number parameters 26.2% decrease floating-point operations. exhibits superior performance aerial detection.

Language: Английский

---

High temporal frequency vehicle counting from low-resolution satellite images

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2023, Volume and Issue: 198, P. 45 - 59

Published: March 8, 2023

Language: Английский

---

A multi-species pest recognition and counting method based on a density map in the greenhouse

Computers and Electronics in Agriculture, Journal Year: 2023, Volume and Issue: 217, P. 108554 - 108554

Published: Dec. 28, 2023

Language: Английский

---

Semantic Segmentation of UAV Images Based on Transformer Framework with Context Information

Mathematics, Journal Year: 2022, Volume and Issue: 10(24), P. 4735 - 4735

Published: Dec. 13, 2022

With the advances in Unmanned Aerial Vehicles (UAVs) technology, aerial images with huge variations appearance of objects and complex backgrounds have opened a new direction work for researchers. The task semantic segmentation becomes more challenging when capturing inherent features global local context UAV images. In this paper, we proposed transformer-based encoder-decoder architecture to address issue precise feature representation is exploited encoder network using self-attention-based transformer framework capture long-range contextual information. A Token Spatial Information Fusion (TSIF) module take advantage convolution mechanism that can details. It fuses details about neighboring pixels makes semantically rich representations. We decoder processes output final level prediction each pixel. demonstrate effectiveness on UAVid Urban Drone datasets, where achieved mIoU 61.93% 73.65%, respectively.

Language: Английский

---

UAV-Assisted Traffic Speed Prediction via Gray Relational Analysis and Deep Learning

Drones, Journal Year: 2023, Volume and Issue: 7(6), P. 372 - 372

Published: June 2, 2023

Accurate traffic prediction is crucial to alleviating congestion in cities. Existing physical sensor-based data acquisition methods have high transmission costs, serious information redundancy, and large calculation volumes for spatiotemporal processing, thus making it difficult ensure accuracy real-time prediction. With the increasing resolution of UAV imagery, use unmanned aerial vehicles (UAV) imagery obtain has become a hot spot. Still, analyzing predicting status after extracting neglected. We develop framework speed extraction based on which consists two parts: module recognition deep learning. First, we learning automate road information, implement vehicle using convolutional neural networks calculate average sections panchromatic multispectral image matching construct dataset. Then, propose an attention-enhanced that considers characteristics increases weights key roads by important fine-grained features twice improve target roads. Finally, validate effectiveness proposed method real data. Compared with baseline algorithm, our algorithm achieves best performance regarding stability.

Language: Английский

---

Remote Sensing Object Counting Through Regression Ensembles and Learning to Rank

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2023, Volume and Issue: 61, P. 1 - 17

Published: Jan. 1, 2023

Remote sensing object counting is finding applications in many fields. Global regression a long-ignored method for counting, though it needs much less manual annotations than the alternatives. This work revisits global and improves two ways—one way by replacing one single regressor with deep ensemble, other breaking down into easier smaller problems: learning to rank (L2R) linear transformation. To this end, we make PAC-Bayesian analysis of ensembles give an upper bound their generalization error, offering new theoretical insight ensemble learning. We also adapt ranking metric optimization scheme suit elegantly handling L2R problem gradient descent. What more, based on our perspective, provide novel building ensembles, which ambiguity constraint imposed. Then, incorporating propose model called "ensemble first-rank-then-estimate networks (eFreeNet)." Our extensive evaluation six benchmarks shows that eFreeNet exhibits compelling performance across board while being more annotation-efficient methods. source code publicly available at https://github.com/huangyongbobo/eFreeNet.

Language: Английский

---

Design and Development of a Novel Sensor Assisted Vehicle Count Prediction Using Modulated Deep Learning Principles

Published: April 4, 2024

Language: Английский

---