Map Information Research Articles

A small‐target traffic sign detection algorithm based on partial conv and atrous spatial pyramid

AbstractTraffic sign detection is essential to an intelligent driving assistance system. The deep learning‐based algorithm proposed in this paper aims to address the issue of low detection accuracy caused by the small size and high density of traffic signs in real‐world traffic scenarios. First, to improve the feature extraction module of the backbone network and to increase the model's ability to capture contextual information, partial convolution (PConv) is introduced. Second, to prevent information loss during the downsampling process, a cross‐stage atrous spatial pyramid (ASPPFCSPC) is constructed using atrous convolution. This method combines feature map information from various scales and expands the receptive field. Lastly, the small‐target detection precision is improved by incorporating an additional small‐target detection head, which uses high‐resolution feature maps for shallow features. The detection head is decoupled to extract the location and class information of the predicted target separately, thereby enhancing the generalization ability of the proposed model. Experiments have demonstrated the superiority of the proposed algorithm, as testing on the TT100K dataset resulted in a mAP@0.5 of 91.2% and a mAP@0.5:0.95 of 71.8% using the proposed algorithm.

Global Path Optimal Planning Method for Autonomous Vehicles Based on Image Feature Extraction

In order to solve the problem of autonomous vehicles driving safely on the road while searching for optimal paths to avoid traffic congestion and obstruction, this paper establishes an optimal path planning model based on image feature extraction. The model consists of four parts: the selection of key turning points, interpolation between turning points, definition of the evaluation function, and global optimal path search. First, based on the map information provided by the network, an image feature detection algorithm was proposed, and the key turning points of the map route were selected and stored in an open table. Then, based on mechanical analysis theory, the cubic interpolation between two key turning points of the optimal path was defined. This allows the car to drive along a continuous curve with a certain arc radius, enabling it to drive smoothly, effectively avoid obstacles, and ensure safety. Then, an evaluation function is defined as needed, and different evaluation functions are specified for different road sections to avoid excessive local deviations in planning and to reduce the impact of known obstacles on the path. Finally, the cost of the search path is calculated based on the evaluation function. According to the principle of minimum cost, a globally optimal heuristic search algorithm is proposed to find the globally optimal path. Experimental results show that the proposed global optimal path planning algorithm has excellent performance, with an average accuracy of 96.71% and a search speed of 2.78 ms. The model not only provides accurate path planning and fast path selection capabilities but also has strong robustness against interference.

Including soil spatial neighbor information for digital soil mapping

Digital soil mapping (DSM) is transforming how we understand and manage soil resources, offering high-resolution spatial–temporal soil information essential for addressing environmental challenges. The integration of environmental covariates has advanced soil mapping accuracy, while the potential of neighboring soil sample data has been largely overlooked. This study introduces soil spatial neighbor information (SSNI) as a novel approach to enhance the predictive power of spatial models. Utilizing two open-access datasets from LUCAS Soil and Meuse, our findings showed that incorporating SSNI improved the accuracy of random forest models in mapping soil organic carbon density (reduced %RMSE of 3.1%), cadmium (reduced %RMSE of 3.6%), copper (reduced %RMSE of 5.9%), lead (reduced %RMSE of 11.5%), and zinc (reduced %RMSE of 7.4%). Compared to the inclusion of buffer distance or oblique geographic coordinates for modelling, SSNI also performed better for both LUCAS Soil and Meuse datasets. This study underscores the value of SSNI in improving digital soil maps by capturing the neighboring information. Embracing SSNI could lead to more informed decision-making in soil management and its potential applicability across other disciplines also remains open for exploration in future research endeavors.

Small Target Detection Algorithm Based on Improved YOLOv5

Small targets exist in large numbers in various fields. They are broadly used in aerospace, video monitoring, and industrial detection. However, because of its tiny dimensions and modest resolution, the precision of small-target detection is low, and the erroneous detection rate is high. Therefore, based on YOLOv5, an improved small-target detection model is proposed. First, in order to improve the number of tiny targets detected while enhancing small-target detection performance, an additional detection head is added. Second, involution is used between the backbone and neck to increase the channel information of feature mapping. Third, the model introduces the BiFormer, wherein both the global and local feature information are captured simultaneously by means of its double-layer routing attention mechanism. Finally, a context augmentation module (CAM) is inserted into the neck in order to maximize the structure of feature fusion. In addition, in order to consider among the required real frame as well as the prediction frame simultaneously, YOLOv5’s original loss function is exchanged. The experimental results using the public dataset VisDrone2019 show that the proposed model has P increased by 13.43%, R increased by 11.28%, and mAP@.5 and mAP@[.5:.95] increased by 13.88% and 9.01%, respectively.

A Pilot Study of Deep Learning-Based Monocular Depth Estimation from Fundus Photographs

The purpose of this study was to evaluate the feasibility of a generalizable deep-learning (DL) based system with no a priori knowledge of fundus photographs to generate monocular depth map information about optic disc structures from this imaging modality. Images of 30 stereo pairs of fundus photographs centered on the optic disc of 30 subjects were analyzed with this DL system to generate monocular depth maps using zero-shot cross-dataset transfer. These maps were registered onto reference standard depth maps derived from Optical Coherence Tomography. Accuracy of the DL system was assessed by the root of mean squared error (RMSE) between the estimate and reference standard. 47% of the total images from the dataset were successfully processed, with mean RMSE of 0.081. Our findings demonstrate that single image, monocular depth estimation with a generalizable DL system using zero-shot cross-dataset transfer applied to retinal color fundus photographs is feasible and has potential.

High-resolution population mapping based on SDGSAT-1 glimmer imagery and deep learning: a case study of the Guangdong-Hong Kong-Marco Greater Bay Area

ABSTRACT Accurate population mapping is crucial for disaster management, urban planning, etc. However, current methods using nighttime light (NTL) and gridded population datasets are limited by low spatial resolution and insufficient training data for complex models such as deep learning. These models do not adequately utilize spatial information in population mapping. To address these limitations, this study proposes a high-resolution population mapping method using the Sustainable Development Goals Science Satellite 1 (SDGSAT-1) glimmer imager data and deep learning. The method includes a sample generation strategy with multiple regression and multilevel screening to provide sufficient, high-quality samples for deep learning. A Fine Population mapping network (FinePop-net) is also developed to train regression models using image samples, capturing multi-scale features for model training. When applied to the Guangdong-Hong Kong-Macao Greater Bay Area with 40-meter resolution SDGSAT 1 glimmer imagery, the method significantly reduced the average absolute error and root-mean-square error by 9.35% and 11.44%, respectively, compared with those of the pixel-level learning methods. It also outperformed other population spatialization datasets and NTL data by over 30% and 10%, respectively, in terms of error reduction. The results highlight the method’s effectiveness and the value of SDGSAT-1 glimmer imagery for fine population spatialization.

A Method for Quantifying Global Network Topology Based on a Mathematical Model

To facilitate a direct comparison of the differences in network resources among countries worldwide, this paper proposes a method for quantifying the relationship between autonomous systems and territorial networks from the perspective of network topology. Using global router-level network topology data as the foundational data for the network resources of various countries, we abstract the dual mapping information of router geographic distribution and operational ownership into a matrix-form mathematical model. By employing relevant indicators from both network scale and border connectivity, we compare matrix model data from different periods to quantitatively assess changes in the network structures of countries globally. The study results show that internet resources are concentrated in the United States, which owns 38.04% of the global routers, distributed across 87.88% of the countries, significantly impacting the global network. Compared to the average quantitative indicators of each country, 67.00% of the countries exhibit higher deployment consistency, 37.30% show higher border connection consistency, 23.81% perform prominently in terms of impact, and 46.20% have outstanding border node degrees. From a continental perspective, the analysis indicates that Asian and African countries have a closer relationship between AS and territorial networks, while Europe’s connections are relatively sparse. Over time, we observe a slight decline in deployment consistency in Asia, Africa, and Europe, a slight increase in border connection consistency in Asia, Africa, and North America, and enhanced impact in Asia, Africa, Europe, and South America. These trends suggest that the integration between AS and territorial networks is intensifying in most countries.

Detection of putative loci affecting milk yield in Turkish Awassi sheep using microsatellite markers.

On the basis of comparisons between bovine and ovine genome mapping information, the aim of the study was to analyze the genetic diversity of selected DNA microsatellites from the bovine genome and to investigate their correlation with the average daily milk yield in Awassi sheep. 18 informative microsatellite markers were selected from the significant QTL regions affecting milk yield identified in the bovine genome in previous studies. The selected microsatellite markers were then amplified by PCR as reciprocal amplifications on the genomic DNA of Awassi sheep, with standard daily milk yield records. Thus, in this study, 18 microsatellite markers associated with milk yield in the bovine genome were examined for both determination of genetic polymorphism within the flock and the effects of marker loci on average daily milk yield in Awassi sheep. Allele frequencies of markers were determined based on the results of fragment analysis. The analysis of variance showed that the 123bp allele at the marker locus BMS1341 on BTA2 significantly influenced the average daily milk yield of Ivesi sheep (P < 0.01). On the other hand, the BMS381 locus with a 115bp allele on BTA2, the MCM140 locus with a 185bp allele on BTA6, the BMS2721 locus with a 155bp allele, the BM1237 locus with 174 and 180bp alleles on BTA7, and finally, the BMS1967 locus with a 117bp allele, the BM4208 locus with 176 and 182bp alleles, and the INRA locus with a185 bp allele on BTA8 showed moderately significant effects on the average daily milk yield of Ivesi ewes (P < 0.05).

Development of the Volcanic Disaster Risk Reduction Education Program Using the ICT Tool “YOU@RISK Volcanic Disaster Edition” —Practical Verification at a Junior High School in the Mt. Nasu Area—

This study developed a volcanic disaster risk reduction education program that aims to help junior high school students visually understand the risks of volcanic eruptions using map information and acquire independent assessments and actions to protect their own lives from eruption disasters. First, this program analyzed the present state, problems, and educational needs of volcanic disaster risk reduction education in Japan, and thereafter designed the learning content to fuse information and communication technology (ICT) education based on the Global and Innovation Gateway for All (GIGA) School Initiative promoted by Japan with disaster risk reduction education from a geographical viewpoint. We adopted the ADDIE model from instructional design theory to develop the program. An important feature of this program is the realization of geography learning using the “YOU@RISK Volcanic Disaster Edition,” a web-based GIS application developed by the National Research Institute for Earth Science and Disaster Resilience (NIED), as an ICT tool. Learners use YOU@RISK on their tablets and perform operations such as visualizing the risk range in the event of a volcanic eruption and searching for appropriate evacuation sites and routes, while acquiring the knowledge and skills necessary for geographical literacy and emergency response actions. The program is able to realize interactive learning, which is a step ahead of education that relies on conventional paper-based hazard maps. Among active volcanoes nationwide, Mt. Nasu was selected as target volcano for this study. In addition, a junior high school located in the area surrounding Mt. Nasu, Nasu Junior High School in Nasu Town, Tochigi Prefecture, was selected as the target school. The program was implemented by teachers using tablets provided for each student at the school. We evaluated the effectiveness of the program by measuring and analyzing the effects of learning through the practice. Consequently, the effectiveness of the volcanic disaster risk reduction education program using ICT was confirmed.

Deep indoor illumination estimation based on spherical gaussian representation with scene prior knowledge

High dynamic range illumination estimation from a single low dynamic range image is a critical task in the fields of computer vision, graphics and augmented reality. However, directly learning the full HDR environment map or parametric lighting information from a single image is extremely difficult and inaccurate. As a result, we propose a two-stage network approach for illumination estimation that integrates spherical gaussian (SG) representation with scene prior knowledge. In the first stage, a convolutional neural network is utilized to generate material and geometric information about the scene, which serves as prior knowledge for lighting prediction. In the second stage, we model indoor environment illumination using 128 SG functions with fixed center direction and bandwidth, allowing only the amplitude to vary. Subsequently, a Transformer-based lighting parameter regressor is employed to capture the complex relationship between the input images with scene prior information and its SG illumination. Additionally, we introduce a hybrid loss function, which combines a masked loss for high-frequency illumination with a rendering loss for improving the visual quality. By training and evaluating the lighting model on the created SG illumination dataset, the proposed method achieves competitive results in both quantitative metrics and visual quality, outperforming state-of-the-art methods.

AVA-YOLO: image-based multiscale feature fusion enhanced perception model for snow avalanche detection

Abstract The global climate change has led to frequent occurrences of snow avalanche disasters. However, the significant variations in scale and shape during the avalanche process, and complex background imagery pose significant challenges to automated detection efforts. There is an urgent need to combine advanced deep learning technology to research automatic detection and recognition of avalanches in the field. In this paper, a novel deep learning model based on YOLOv8 improved multi-scale detection called AVA-YOLO is proposed to solve this problem. In AVA-YOLO, a key component, AKA (AKConv Combined Attention) module was designed and developed. This module combines the deformable convolutional properties of AKConv with the state-of-the-art self-attention module Exponential Moving Average, aiming to better perceive the feature map information of different shaped avalanches and to enhance the global relevance, thus improving the utilization of the information. Secondly, a new multi-scale sensing network structure was designed by increasing the number of detection heads to four and introducing the AKA module into the key positions of the network, while the association between model layers was newly designed to enhance the fusion of shallow and deep information to improve the detection accuracy. Experimental results demonstrated the effectiveness of AVA-YOLO, achieving 95.7% mAP50 and 75.6% mAP50:95 detection accuracies, as well as an F1 score of 0.92. Finally, a number of experiments were conducted to demonstrate the superior performance of the proposed model in comparison to other versions of YOLO, which will further exploit the potential of webcams as an underutilized technical capability in snow avalanche intelligence and portable monitoring.

Investigative power of genomic informational field theory relative to genome-wide association studies for genotype-phenotype mapping.

Identifying associations between phenotype and genotype is the fundamental basis of genetic analyses. Inspired by frequentist probability and the work of R. A. Fisher, genome-wide association studies (GWAS) extract information using averages and variances from genotype-phenotype datasets. Averages and variances are legitimated upon creating distribution density functions obtained through the grouping of data into categories. However, as data from within a given category cannot be differentiated, the investigative power of such methodologies is limited. Genomic informational field theory (GIFT) is a method specifically designed to circumvent this issue. The way GIFT proceeds is opposite to that of GWAS. Although GWAS determines the extent to which genes are involved in phenotype formation (bottom-up approach), GIFT determines the degree to which the phenotype can select microstates (genes) for its subsistence (top-down approach). Doing so requires dealing with new genetic concepts, a.k.a. genetic paths, upon which significance levels for genotype-phenotype associations can be determined. By using different datasets obtained in Ovis aries related to bone growth (dataset 1) and to a series of linked metabolic and epigenetic pathways (dataset 2), we demonstrate that removing the informational barrier linked to categories enhances the investigative and discriminative powers of GIFT, namely that GIFT extracts more information than GWAS. We conclude by suggesting that GIFT is an adequate tool to study how phenotypic plasticity and genetic assimilation are linked.NEW & NOTEWORTHY The genetic basis of complex traits remains challenging to investigate using classic genome-wide association studies (GWASs). Given the success of gene editing technologies, this point needs to be addressed urgently since there can only be useful editing technologies whether precise genotype-phenotype mapping information is available initially. Genomic informational field theory (GIFT) is a new mapping method designed to increase the investigative power of biological/medical datasets suggesting, in turn, the need to rethink the conceptual bases of quantitative genetics.

Hyper-relational interaction modeling in multi-modal trajectory prediction for intelligent connected vehicles in smart cites

Trajectory prediction of surrounding traffic participants is vital for the driving safety of Intelligent Connected Vehicles (ICVs). It has been enabled with the help of the availability of multi-sensor information collected by ICVs. For accurately predicting the future movements of traffic agents, it is crucial to subtlety model the inter-agent interaction. However, existing works focus on the correlations between agents and the map information while neglecting the importance of directly modeling the impact of map elements on inter-agent interactions, the direct modeling of which is beneficial for the representation of agent behaviors. Against this background, we propose to model the hyper-relational interaction, which incorporates map elements into the inter-agent interaction. To tackle the hyper-relational interaction, we propose a novel Hyper-relational Multi-modal Trajectory Prediction (HyperMTP) approach. Specifically, a hyper-relational driving graph is first constructed and the hyper-relational interaction is represented as the hyperedge, directly connecting to various nodes (i.e., agents and map elements). Then a structure-aware embedding initialization technique is developed to obtain unbiased initial embeddings. Afterward, hypergraph dual-attention networks are designed to capture correlations between graph elements while retaining the hyper-relational structure. Finally, a heterogeneous Transformer is devised to further capture the correlations between agents’ states and their corresponding hyper-relational interactions. Experimental results show that HyperMTP consistently outperforms the best-performing baseline with an average improvement of 4.8% across two real-world datasets. Moreover, HyperMTP also boosts the interpretability of trajectory prediction by quantifying the impact of map elements on inter-agent interactions.

Spectrum Sensing Method Based on STFT-RADN in Cognitive Radio Networks.

To address the common issues in traditional convolutional neural network (CNN)-based spectrum sensing algorithms in cognitive radio networks (CRNs), including inadequate signal feature representation, inefficient utilization of feature map information, and limited feature extraction capabilities due to shallow network structures, this paper proposes a spectrum sensing algorithm based on a short-time Fourier transform (STFT) and residual attention dense network (RADN). Specifically, the RADN model improves the basic residual block and introduces the convolutional block attention module (CBAM), combining residual connections and dense connections to form a powerful deep feature extraction structure known as residual in dense (RID). This significantly enhances the network's feature extraction capabilities. By performing STFT on the received signals and normalizing them, the signals are converted into time-frequency spectrograms as network inputs, better capturing signal features. The RADN is trained to extract abstract features from the time-frequency images, and the trained RADN serves as the final classifier for spectrum sensing. Experimental results demonstrate that the STFT-RADN spectrum sensing method significantly improves performance under low signal-to-noise ratio (SNR) conditions compared to traditional deep-learning-based methods. This method not only adapts to various modulation schemes but also exhibits high detection probability and strong robustness.

Vehicle modeling and state estimation for autonomous driving in terrain

The automobile industry usually ignores the height of the path and uses planar vehicle models to implement automatic vehicle control. In addition, existing literature mostly concerns level terrain or homogeneous road surfaces for estimating vehicle dynamics. However, ground vehicles utilized in forestry, such as forwarders, operate on uneven terrain. The vehicle models built on level terrain assumptions are inadequate to capture the rolling or pitching dynamics of such machines as rollover of such vehicles is a potential risk. Therefore, knowledge about the height profile of the path is crucial for automating such off-road operations and avoiding rollover. We propose the use of a six-degrees-of-freedom (6-DOF) dynamic vehicle model to solve the autonomous forwarder problem. An adaptive linear tire model is used in the 6-DOF model assuming the vehicle operates in a primary handling regime. The force models are modified to include the three-dimensional (3D) map information. The calibration procedures, identifying actuator dynamics, and quantifying sensor delays are also represented.The proposed vehicle modeling contributed to realizing the continuous-discrete extended Kalman filter (CDEKF), which takes into account the 3D path during filtering and fixed-lag smoothing. Polaris (an all-terrain electric car) is used as a case study to experimentally validate the vehicle modeling and performance of the state estimator. Three types of grounds are selected — an asphalt track, a concrete track with a high elevation gradient, and a gravel track inside a forest. Stable state estimates are obtained using CDEKF and sparse 3D maps of terrains despite discontinuities in satellite navigation data inside the forest. The height estimation results are obtained with sufficient accuracy when compared to ground truth obtained by aerial 3D mapping. Finally, the proposed model’s applicability for predictive control is demonstrated by utilizing the state estimates to predict future states considering (3D) terrain.

Information Professionals’ Metaphorical Perceptions of Artificial Intelligence Concept

This study examines the metaphors used by 66 information professionals in Türkiye to conceptualize artificial intelligence (AI). Through qualitative metaphor analysis, rich insights emerged on perceptions of AI’s nature, functioning, relationships with humans, roles, and societal impacts. Some of the metaphors used are child, human, artist, human intelligence, robot, assistant, doctor, weapon, terminator, something scary, closed box, black hole, Pandora’s box, etc. Multiple, often contradictory metaphors like “child” and “uncontrollable power” indicate AI is viewed in a multidimensional way, evoking both promise and concern. Key findings suggest information professionals appreciate AI’s transformative potential in enhancing services but also harbor anxieties related to human relevance, technocracy, and social issues like privacy. Understanding these complex cognitive and emotional orientations can facilitate responsible AI adoption to augment information services. Gesturing at AI’s multifaceted connotations also foregrounds communication challenges complicating public debate and policymaking on emerging technologies. The value of this study lies in surfacing the nuanced perspectives of a stakeholder group positioned to mediate public understanding and direct integration of AI in information services. Mapping information professionals’ rich construals of AI also highlights the need for public communication attuned to metaphorical reasoning around new technologies.

Program context-assisted address translation for high-capacity SSDs

As the capacity of NAND flash-based SSDs keeps increasing, it becomes crucial to design a memory-efficient address translation algorithm that offers high performance when a translation table cannot be entirely loaded in a controller DRAM. Existing flash translation layers (FTL) employ demand-based address translation which caches popular mapping information in DRAM by leveraging locality of I/O references. Owing to the lack of information about detailed behaviors of applications, however, existing demand-based FTLs often suffer from many translation-table misses and thus result in sub-optimal performance. In this paper, we propose a new Program context-AssisteDFlash Translation Layer, called PADFTL. Unlike existing FTLs which are implemented as the form of firmware, PADFTL is vertically integrated with a host-level I/O classifier which provides useful hints for an FTL in an SSD to make a better decision in managing a translation table. The host-level I/O classifier monitors unique behaviors of applications by analyzing their program contexts and categorizes I/O patterns into four types, (1) Loop, (2) Hot, (3) Sequential, and (4) Random, which are then delivered to an SSD through extended interfaces. The SSD-side module of PADFTL partitions a controller DRAM into four zones and isolates mapping information associated with different I/O patterns into separate zones. By employing cache management strategies optimized for individual zones, PADFTL can lower the overall translation-table miss ratio. To evaluate the effectiveness of PADFTL, we implement the host-level classifier in the Linux kernel and PADFTL’s FTL in a trace-driven FTL simulator. In our experimental results, compared to the state-of-the-art FTL, PADFTL increases the overall table hit ratio by 16% while reducing the address translation time by up to 20% on average.

A novel EMS design framework for SPHTs based on instantaneous layer, driving event layer, and driving cycle layer

To effectively harness the energy-saving potential of series–parallel hybrid transmissions (SPHTs) with multiple gears and modes and to enhance the driving cycle adaptability of the rule-based energy management strategy (RB EMS), this study establishes a novel EMS design framework for SPHTs at three levels: the instantaneous, the driving event, and the driving cycle layers. Firstly, an equivalent fuel factor based on the principles of energy conservation and calorific value measurements is constructed, the energy consumption of different working modes and torque combinations under different instantaneous power units is determined, and the decision conditions of working modes are determined. Secondly, the energy consumption of engine and motor torque combination under constant speed, acceleration and deceleration driving events is compared, and the torque distribution feature of multi-power sources is determined. Thirdly, a driving cycle distribution factor is introduced and calculated using vehicle navigation map information, which can adjust the gear switching condition of parallel driving mode online. Based on above, an RB EMS adjusted with driving cycles (ADC-RB EMS) is proposed. The effectiveness of the proposed strategy is then verified through real-world vehicle tests. The fuel consumption performance of this strategy falls between the RB EMS and the dynamic programming (DP) strategy. According to the fuel consumption results, the RB EMS consumes an additional 0.35 L/100 km of fuel, while the DP strategy saves only 3.66% more energy compared to this strategy. The application potential of driving cycle distribution factor in instantaneous optimal control is tested. Compared with ADC-RB and CD-CS, the fuel saving rate is 1.71% and 5.67% respectively, which proves the energy saving performance of A-ECMS. This study provides a development direction for improving the energy saving effect of the RB EMS.

Optimizing Spatial Accuracy for Photogrammetric Processing of Drone-based 3D Mapping

Unmanned aerial vehicle (UAV) systems have become crucial for gathering information for observation, surveillance, mapping, and 3D modeling tasks. The use of UAVs in close and mid-range sectors has shown potential for cost-effective alternatives to traditional aerial photogrammetry conducted by humans. The research aims to optimize photogrammetric processing for drone-based 3D mapping by examining strategies and applications to increase accuracy and efficiency. The study highlights the significance of image overlap and high-quality camera equipment to obtain reliable outcomes for stereo-photogrammetry and depth measurements. Drones and specialized software like 3DF Zephyr and Pix4D were used to create a 3D map. The software uses automatic structure from motion techniques, including feature extraction, image matching, and bundle block adjustment, to produce a dense point cloud that forms the basis for the 3D model. A DGPS system was implemented to enhance spatial accuracy. The map initially showed inadequate accuracy, with an error rate exceeding 80cm. However, with the use of a DGPS system, the error was reduced to less than 3cm. This study provides suggestions and insights for improving photogrammetric processing for drone-based 3D mapping.

A dynamic object removing 3D reconstruction system based on multi-sensor fusion

Abstract Currently, one of the key technologies for autonomous navigation of unmanned mobile robots is SLAM, which faces many challenges in practical applications. These challenges include a lack of texture, deterioration in sensor performance, and interference from moving objects in dynamic outdoor environments, all of which have an impact on the mapping system. To address these issues, this paper proposes a framework for lidar, vision camera, and inertial navigation data, resulting in fusion and dynamic object removing. The system consists of three sub-modules: the Lidar-Inertial Module (LIM), the Visual-Inertial Module (VIM), and the Dynamic-Object-Removing Module (DORM). LIM and VIM assist each other, with lidar point clouds providing three-dimensional information for the global voxel map and the camera providing pixel-level color information. At the same time, the DORM performs synchronous dynamic object detection to remove dynamic objects from the global map. The system constructs a multi-sensor factor graph using the state and observation models, and the optimal solution is obtained using least squares. Furthermore, this paper employs triangle descriptors and bundle adjustment methods for loop closure detection in order to reduce accumulated errors and maintain consistency. Experimental results demonstrate that the system can perform clean state estimation, dynamic removing and scene reconstruction in a variety of complex scenarios.