Matching Position Research Articles

Evaluation of Deep Learning Models in Search by Example using Capsule Endoscopy Images

Wireless capsule endoscopy (WCE) has revolutionized the field of gastrointestinal examinations, being MedtronicTM WCE one of the most used in clinics. In those WCE videos, medical experts use RAPID READERTM tool to annotate findings in videos. However, the frame annotations are not available in an open format and, when exported, they have different resolutions and some annotated artefacts that make difficult their localization in the original videos. This difficult the use of WCE medical experts’ annotations in the research of new computed-aid diagnostic (CAD) methods. In this paper, we propose a methodology to compare image similarities and evaluate it in a private MedtronicTM WCE SB3 video dataset to automatically identify the annotated frames in the videos. We used state-of-the-art pre-trained convolutional neural network (CNN) models, including MobileNet, InceptionResNetv2, ResNet50v2, VGG19, VGG16, ResNet101v2, ResNet152v2, and DenseNet121, as frame features extractors and compared them with the Euclidean distance. We evaluated the methodology performance on a private dataset consisting of 100 WCE videos, totalling 905 frames. The experimental results showed promising performance. The MobileNet model achieved an accuracy of 94% for identifying the first match, while the top 5, top 10, and top 20 matches were identified with accuracies of 94%, 94%, and 98%, respectively. The VGG16 and ResNet50v2 models also demonstrated strong performance, achieving accuracies ranging from 88% to 93% for various match positions. These results highlight the effectiveness of our proposed methodology in localizing target frames and even identifying similar frames very use useful for training data-driven models in CAD research. The code utilized in this experiment is available on the Github†.

Integrated Shared Random Key Agreement Protocol for Wireless Sensor Network

The secured data transmission in Wireless Sensor Network (WSN) relies on effective key generation and secured sharing. The generated key must be random to enhance data confidentiality. The processes associated with the security in WSN must be designed at reduced computing time and communication cost. Our research work aims to design a novel lightweight key-sharing protocol that is needed for ensuring data confidentiality. The protocol must meet the constraints of WSN by being lightweight and consuming less energy. Security breaches in WSNs occur due to insecure keys. This can be overcome by generating shared keys which are generated once using the dynamic features of Sensor Nodes (SNs) when the Cluster Heads (CHs) are selected. In this research work, we have generated the Master Shared Key (MSK) at the transmitter node by forming a Galois Ring (GR) using WSN parameters and derived the Shared Random Key (SRK) using matched positions of exchanged Random Sequences (RSs). It is protected using a Physically Unclonable Function (PUF). The novelty lies in the SRK generation from MSK which is chosen at random from the polynomials generated during the formation of GR. The MSK is securely shared with the receiver node by encrypting using a Preloaded Key (PK). After this exchange, the key for encryption and decryption is derived by the transmitter and the receiver by exchanging RSs. The SRK is then encrypted using a key which is a unique fingerprint of the SN generated using PUF and stored in the SNs and the CHs to prevent node capture attack that occurs in WSN. Our proposed Shared Random Key Agreement Protocol (SRKAP) is comparable to the Localized Encryption and Authentication Protocol (LEAP) and performs better compared to the Elliptic Curve Diffie Hellman (ECDH) algorithm

White matter disconnection impacts proprioception post-stroke.

Proprioceptive impairments occur in approximately 50-64% of people following stroke. While much is known about the grey matter structures underlying proprioception, our understanding of the white matter correlates of proprioceptive impairments is less well developed. It is recognised that behavioural impairments post-stroke are often the result of disconnection between wide-scale brain networks, however the disconnectome associated with proprioception post-stroke is unknown. In the current study, white matter disconnection was assessed in relation to performance on a robotic arm position matching (APM) task. Neuroimaging and robotic assessments of proprioception were collected for 203 stroke survivors, approximately 2-weeks post-stroke. The robotic assessment was performed in a KINARM Exoskeleton robotic device and consisted of a nine-target APM task. First, the relationship between white matter tract lesion load and performance on the APM task was assessed. Next, differences in the disconnectome between participants with and without impairments on the APM task were examined. Greater lesion load to the superior longitudinal fasciculus (SLF II and III), arcuate fasciculus (all segments) and fronto-insular tracts were associated with worse APM task performance. In those with APM task impairments, there was, additionally, disconnection of the posterior corpus callosum, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus and optic radiations. This study highlights an important perisylvian white matter network supporting proprioceptive processing in the human brain. It also identifies white matter tracts, important for relaying proprioceptive information from parietal and frontal brain regions, that are not traditionally considered proprioceptive in nature.

Discharge characteristics of 100-W Hall thrusters under position matching between the anode and zero-magnetic-field region

Poor performance limits the wider application of 100-W-power class Hall thrusters. The effects of position matching between the anode and the zero-magnetic-field region (ZMFR) on the discharge process and performance improvement of 100-W Hall thrusters were investigated. The experimental results demonstrate that appropriately widening the ZMFR downstream of the anode improves the ion current and thrust, promotes plume focusing, and optimizes thruster performance. The thrust and anode efficiency is enlarged by 7.6 % and 20 % (relative value), respectively, when the ZMFR downstream of the anode is expanded by 0.2L (L is the channel length). The simulation results indicated that widening the ZMFR promoted the expansion of the high-ionization zone to the bottom of the channel, and increased the overall ionization rate. Moreover, as the ZMFR increases, the effective ion acceleration voltage increases. However, widening the ZMFR also aggravated the ion-wall interaction and reduced the propellant utilization rate. Therefore, the performance of the thrusters exhibited an optimal range as the length of the ZMFR increased. In addition, increasing the ZMFR did not affect the life of the thrusters.

​​ANALYSIS AND OPTIMIZATION OF COLLEGE STUDENT'S EDUCATION AND TRAINING BASED ON TALENT MARKET DEMAND​

Research shows that there is a problem of low employment happiness index among Chinese college graduates in employment. Through literature research and market survey, we found that geographical differences and low matching of positions and majors are the main reasons. Therefore, we deeply analyze the logical relationship between them and propose improvement suggestions and directions for the above problems. This study aims to examine the correlation between school discipline settings, education methods and employment from the actual needs of society, so as to design a feedback mechanism to optimize the dependent variable through the inverse adjustment of the independent variable and finally solve the above-mentioned employment happiness problems. The samples for this study were collected from fresh graduates of universities in Guangdong Transportation Vocational College and several nearby cities. The research questionnaire was designed by the researcher himself and issued to and solicited from these graduates and school staff. The results showed that regional economic differences had little effect on graduates' employment well-being index, but education patterns and curriculum had a greater effect on graduates' employment well-being index. These findings have important theoretical, pedagogical, and practical implications.

Research on Mining Talent Demand for E-commerce Majors based on LDA Topic Model

[Purpose] Based on the text mining method, this study analyzes the job demand for domestic e-commerce industry positions in the Internet-oriented recruitment data, promotes the matching of e-commerce positions and talents, and promotes the good construction of the employment environment in the domestic e-commerce industry.[Method] Use the LDA(Latent dirichlet allocation) topic model for job competency requirements to calculate the similarity of the word segmentation results for professional talent requirements, determine the optimal number of topics, and output the visualization results of the LDA topic model.[Conclusion] In eastern China, there is a high demand for e-commerce jobs, and the majority of these positions only call for 1-3 years of experience. The basic requirement for the majority of jobs is a college degree. The Director of Operations and Director of Network Operations focus on e-commerce operations and management, the Network Promotion Specialist focuses on e-commerce advertising and product promotion, the Data Analyst and Market Analyst focus on data analysis and market research, and the Sales Representative focuses on sales. There is a correlation between the salary of the positions and education, experience and region; the higher the education, the more experience and the higher the level of regional development, the higher the salary level.

Optical Phased Array-Based Laser Beam Array Subdivide Pixel Method for Improving Three-Dimensional Imaging Resolution

The small number of pixels in the current linear mode avalanche photodiode (LM-APD) array limits its three-dimensional (3D) imaging resolution. We use an optical phased array-based beam array subdivided pixel method to improve the 3D imaging resolution, using an optical phased array to generate a beam array with the same number of pixels as the LM-APD array and matching positions and controlling each sub-beam in the beam array to scan in the field of view of the corresponding pixel. The sub-beam divergence angle in the beam array is smaller than the instantaneous field of view angle of a single pixel in the LM-APD array. The sub-beam scanning in a single pixel’s field of view realizes the multiple acquisition of the target 3D information by the LM-APD array, thus improving the resolution of the LM-APD array. The distribution of the beam array in the far field is simulated, and the main performance parameters of 3D imaging are analyzed. Finally, a liquid crystal phase modulator is used as an optical phased array device to conduct experiments on a target 20 m away, and the results prove that our method can improve the resolution from 4 × 4 to 8 × 8, which can be improved at least four times.

PPO-Based Joint Optimization for UAV-Assisted Edge Computing Networks

In next-generation mobile communication scenarios, more and more user terminals (UEs) and edge computing servers (ECSs) are connected to the network. To ensure the experience of edge computing services, we designed an unmanned aerial vehicle (UAV)-assisted edge computing network application scenario. In the considered scenario, the UAV acts as a relay node to forward edge computing tasks when the performance of the wireless channel between UEs and ECSs degrades. In order to minimize the average delay of edge computing tasks, we design the optimization problem of joint UE–ECS matching and UAV three-dimensional hovering position deployment. Further, we transform this mixed integer nonlinear programming into a continuous-variable decision process and design the corresponding Proximal Policy Optimization (PPO)-based joint optimization algorithm. Sufficient data pertaining to latency demonstrate that the suggested algorithm can obtain a seamless reward value when the number of training steps hits three million. This verifies the algorithm’s desirable convergence property. Furthermore, the algorithm’s efficacy has been confirmed through simulation in various environments. The experimental findings ascertain that the PPO-based co-optimization algorithm consistently attains a lower average latency rate and a minimum of 8% reduction in comparison to the baseline scenarios.

Dimension-Expanded-Based Matching Method With Siamese Convolutional Neural Networks for Gravity-Aided Navigation

Matching algorithm is the key technique of the gravity-aided inertial navigation system. With the development of artificial intelligence, many neural network-based matching methods have been extensively studied. The pattern recognition-based matching methods transforms the matching problem as pattern recognition, which cannot be used directly on datasets where the neural networks have not been trained. To improve the accuracy of navigation and positioning, it is necessary to extract muti-dimensional gravity features from the limited navigation information. In this paper, the sequence of the gravity anomaly value is expanded to two-dimensional feature map containing time series features by Gramian Angular Fields method, which preserves the numerical information of the one-dimensional sequence and extracts the correlation relationship between each element. In addition, to reduce the influence of gravity measurement instrument error on the position precision of gravity matching algorithm, affine transformation is performed on INS trajectory and a Siamese convolutional neural network model is proposed to compare the measured gravity database with the gravity anomaly value in the pre-stored gravity background map and get the matching position. Simulation results and practical tests show that the proposed method can obtain a more precise location result compared with the traditional matching algorithm.

Comprehensive Image Matching Algorithm Based on Local GLCM for Gravity-Gradient-Aided Navigation

Matching algorithm is the key technology of gravity-gradient-aided inertial navigation system (INS). Traditional single-point and sequence matching algorithms fail to make full use of the multicomponent gravity gradient reference map information, so that the matching accuracy cannot meet the requirements. A gravity gradient comprehensive image matching algorithm based on local gray-level co-occurrence matrix (GLCM) is proposed. Gravity gradient reference maps are prepared by using the frequency-domain Fourier transform method. Gravity gradient real-time images are constructed within the confidence interval of INS. The state equation and observation equation are established. The optimal matching image is determined by synthesizing the gravity gradients in five independent directions through the similarity coarse screening and local GLCM feature matching. Finally, the matching position coordinates are calculated in reverse. Simulation and experimental results show that the matching accuracy of the proposed algorithm is within one grid in both short-term and long-term matching errors. It is proved that the proposed algorithm not only has high accuracy and better robustness, but also has no strict requirements on the change of gravity gradient characteristics in the navigation area.

86 Unfilled in Emergency Medicine: An Analysis of Open 2022 and 2023 Match Positions by Program Accreditation, Ownership, and Geography

86 Unfilled in Emergency Medicine: An Analysis of Open 2022 and 2023 Match Positions by Program Accreditation, Ownership, and Geography

Initial clinical evaluation of a novel combined biometric, radio-frequency identification, and surface imaging system

PurposeTo evaluate and characterize the overall clinical functionality and workflow of the newly released Varian Identify system (version 2.3). MethodsThree technologies included in the Varian Identify system were evaluated: patient biometric authentication, treatment accessory device identification, and surface-guided radiation therapy (SGRT) function. Biometric authentication employs a palm vein reader. Treatment accessory device verification utilizes two technologies: device presence via Radio Frequency Identification (RFID) and position via optical markers. Surface-guidance was evaluated on both patient orthopedic setup at loading position and surface matching and tracking at treatment isocenter. A phantom evaluation of the consistency and accuracy for Identify SGRT function was performed, including a system consistency test, a translational shift and rotational accuracy test, a pitch and roll accuracy test, a continuous recording test, and an SGRT vs Cone-Beam CT (CBCT) agreement test. Results201 patient authentications were verified successfully with palm reader. All patient treatment devices were successfully verified for their presences and positions (indexable devices). The patient real-time orthopedic pose was successfully adjusted to match the reference surface captured at simulation. SGRT-reported shift consistency against couch readout was within (0.1 mm, 0.030). The shift accuracy was within (0.3 mm, 0.10). In continuous recording mode, the maximum variation was 0.2 ± 0.12 mm, 0.030 ± 0.020. The difference between Identify SGRT offset and CBCT was within (1 mm, 10). ConclusionsThis clinical evaluation confirms that Identify accurately functions for patient palm identification and patient treatment device presence and position verification. Overall SGRT consistency and accuracy was within (1 mm, 10), within the 2 mm criteria of AAPM TG302.

The impact of virtual interviews on the Fellowship Council application and main match process.

The COVID-19 pandemic led the Fellowship Council (FC) to transition rapidly from in-person to virtual interviews. We investigated the impact of this transition on the FC application and main match process. Five years (2018-2022) of deidentified FC applicant, program, and match rank data were used to assess differences between in-person (2018-2019) and virtual interview (2021-2022) cycles. Data are expressed as mean ± SD and one-way and two-way MANOVA tests were applied. Trainees applied to an average of 30.4 ± 24.3 programs and ranked an average of 10.7 ± 9.7 programs with a 57% match rate and average rank position of 3.6 ± 3.3. Fellowship programs received an average of 64.9 ± 28.6 applications and ranked an average of 15.4 ± 8.8 applicants with a 95% match rate and average applicant rank position of 3.0 ± 3.4. Applicants who interviewed virtually applied to a greater number of programs (32.7 vs. 27.0; p < 0.001) and ranked a greater number of programs (11.5 vs. 10.0; p = 0.004) with no difference in match rates (58% vs. 55%, p = 0.291). Among matched applicants, there was a significant difference in average rank position (3.20 vs. 4.30, p < 0.001), favoring the in-person cohort. Fellowship programs had more applicants per program (69.2 vs. 57.8; p < 0.001) and ranked more applicants (17.4 vs. 13.3; p < 0.001) during the virtual interview cycles. No difference in either match rates (93% vs. 96%, p = 0.178) or applicant rank position (3.09 vs. 2.93, p = 0.561) was seen between in-person and virtual application cycles. Virtual interviews were associated with an increased number of applications for fellowship and applicants ranked by programs but did not impact match rates of either group. Rank match position declined somewhat for applicants but not for fellowship programs. Virtual interviews offer more opportunities for applicants and a greater number of candidates for fellowship programs with only a slight decrement in fellow match rank position.

Tracer particle motion driven by vortex formation in the bottom boundary layer underneath internal solitary waves

Internal solitary waves (ISWs) of large amplitude moving in the coastal ocean induce sizeable horizontal velocities above the sea bed. In turn, these give rise to instability and vortex formation in the bottom boundary layer (BBL), and sediment resuspension and concentration maintenance in the water column. We present two-dimensional laminar simulations in a numerical tank suitable for internal wave motion, including the processes of the BBL. The combined wave and vorticity field encounters a cloud of tracer particles near the bottom. The tracer particles are moved vertically because of the vorticity field during a first encounter. The reflected wave intercepts a second time with the tracer particles, which are then moved further vertically. Numerical experiments with a kinematic viscosity of 1/100 cm2 s−1 or 1/1000 cm2 s-1 are used to manipulate the scale of the Reynolds number at a moderate and great laboratory scale. The final vertical position of the tracer particles is found below a vertical level of approximately 0.23 times the water depth (H) after the second passage. The result is independent of the scale. This vertical position matches available field measurements of a summer benthic nepheloid layer reaching a height of 0.19H. The laminar model predictions compare very well to the ISW-driven vortex formation measured in a three-dimensional laboratory wave tank. Convergence of the calculated vortex formation is documented.

Accurate design of spatially separated double active site in Bi4NbO8Cl single crystal to promote Z - Scheme photocatalytic overall water splitting

The efficiency of photocatalytic overall water splitting was mainly limited by the slow reaction kinetics of water oxidation. How to design effective surface active site to overcome the slow water oxidation reaction was a major challenge. Here, we propose a strategy to accelerate surface water oxidation through the fabrication spatially separated double active sites. FeCoPi/Bi4NbO8Cl-OVs photocatalyst with spatially separated double active site was prepared by hydrogen reduction photoanode deposition method. Due to the high matching of the spatial loading positions of FeCoPi and OVs with the photogenerated charge distribution of Bi4NbO8Cl and corresponding reaction mechanisms of substrate, the FeCoPi and OVs on the (001) and (010) crystal planes of Bi4NbO8Cl photocatalyst provided surface active site for water oxidation reaction and electron shuttle reaction (Fe3+/Fe2+), respectively. Under visible light irradiation, the evolution O2 rate of FeCoPi/Bi4NbO8Cl OVs was 16.8 μmol h−1, as 32.9 times as Bi4NbO8Cl. Furthermore, a hydrogen evolution co-catalyst PtRu@Cr2O3 was prepared by sequential photodeposition method. Due to the introduction of Ru, the Schottky barrier between PbTiO3 and Pt was effectively reduced, which promoted the transfer of photogenerated electrons to PtRu@Cr2O3 thermodynamically, the evolution H2 rate on PtRu@Cr2O3/PbTiO3 increased to 664.8 times. On based of the synchronous enhancement of the water oxidation performance on FeCoPi/Bi4NbO8Cl-OVs and water reduction performance on PtRu@Cr2O3/PbTiO3, a novel Z - Scheme photocatalytic overall water splitting system (FeCoPi/Bi4NbO8Cl-OVs) mediated by Fe3+/Fe2+ had successfully constructed. Under visible light irradiation, the evolution rates of H2 and O2 were 2.5 and 1.3 μmol h−1, respectively. This work can provide some reference for the design of active site and the controllable synthesis of OVs spatial position. On the other hand, the hydrogen evolution co catalyst (PtRu@Cr2O3) and the co catalyst FeCoPi for oxygen evolution contributed to the construction of an overall water splitting system.

Effects of Biofeedback on Muscle Effort Reduction when Holding Positions with a Powered Upper Limb Exoskeleton

Exoskeletons have the potential to support daily activities by assisting the movement performance. Previous studies have shown that powered elbow exoskeletons can reduce muscle effort during continuous cyclic movement. However, natural movements have embedded stationary hold periods as well as transitions between postures. The human performance when using exoskeletons for more natural movements should be evaluated. In this study, we implemented visual and haptic electromyography (EMG) biofeedback to help people use an upper limb exoskeleton to perform a target position matching task. Participants (n=36) did not significantly reduce their muscle effort during hold periods when provided with biofeedback. Participants had difficulty relaxing their muscles at more flexed postures during hold periods, suggesting that they continued to provide effort instead of taking advantage of the device. To fully benefit from robotic exoskeletons, additional training and more advanced controllers might be needed.

2chADCNN: A Template Matching Network for Season-Changing UAV Aerial Images and Satellite Imagery

Visual navigation based on image matching has become one of the most important research fields for UAVs to achieve autonomous navigation, because of its low cost, strong anti-jamming ability, and high performance. Currently, numerous positioning and navigation methods based on visual information have been proposed for UAV navigation. However, the appearance, shape, color, and texture of objects can change significantly due to different lighting conditions, shadows, and surface coverage during different seasons, such as vegetation cover in summer or ice and snow cover in winter. These changes pose greater challenges for feature-based image matching methods. This encouraged us to overcome the limitations of previous works, which did not consider significant seasonal changes such as snow-covered UAV aerial images, by proposing an image matching method using season-changing UAV aerial images and satellite imagery. Following the pipeline of a two-channel deep convolutional neural network, we first pre-scaled the UAV aerial images, ensuring that the UAV aerial images and satellite imagery had the same ground sampling distance. Then, we introduced attention mechanisms to provide additional supervision for both low-level local features and high-level global features, resulting in a new season-specific feature representation. The similarity between image patches was calculated using a similarity measurement layer composed of two fully connected layers. Subsequently, we conducted template matching to estimate the UAV matching position with the highest similarity. Finally, we validated our proposed method on both synthetic and real UAV aerial image datasets, and conducted direct comparisons with previous popular works. The experimental results demonstrated that our method achieved the highest matching accuracy on multi-temporal and multi-season images.

Clinical, Neuroimaging and Robotic Measures Predict Long-Term Proprioceptive Impairments following Stroke.

Proprioceptive impairments occur in ~50% of stroke survivors, with 20-40% still impaired six months post-stroke. Early identification of those likely to have persistent impairments is key to personalizing rehabilitation strategies and reducing long-term proprioceptive impairments. In this study, clinical, neuroimaging and robotic measures were used to predict proprioceptive impairments at six months post-stroke on a robotic assessment of proprioception. Clinical assessments, neuroimaging, and a robotic arm position matching (APM) task were performed for 133 stroke participants two weeks post-stroke (12.4 ± 8.4 days). The APM task was also performed six months post-stroke (191.2 ± 18.0 days). Robotics allow more precise measurements of proprioception than clinical assessments. Consequently, an overall APM Task Score was used as ground truth to classify proprioceptive impairments at six months post-stroke. Other APM performance parameters from the two-week assessment were used as predictive features. Clinical assessments included the Thumb Localisation Test (TLT), Behavioural Inattention Test (BIT), Functional Independence Measure (FIM) and demographic information (age, sex and affected arm). Logistic regression classifiers were trained to predict proprioceptive impairments at six months post-stroke using data collected two weeks post-stroke. Models containing robotic features, either alone or in conjunction with clinical and neuroimaging features, had a greater area under the curve (AUC) and lower Akaike Information Criterion (AIC) than models which only contained clinical or neuroimaging features. All models performed similarly with regard to accuracy and F1-score (>70% accuracy). Robotic features were also among the most important when all features were combined into a single model. Predicting long-term proprioceptive impairments, using data collected as early as two weeks post-stroke, is feasible. Identifying those at risk of long-term impairments is an important step towards improving proprioceptive rehabilitation after a stroke.

No relationship between proprioception and interoception in university athletes and non-athletes

Proprioception and interoception are two complex physiological functions believed to be dependent on a variety of factors. The present study examined the relationship between proprioception and interoception in a sample of undergraduate students to determine whether varsity athlete status moderated the relationship, as physical activity levels have been hypothesized to influence both proprioception and interoception. Participants (n = 24) completed two tasks: a heartbeat discrimination test of interoceptive accuracy and a joint position matching test of proprioceptive accuracy at the elbow joint. Participants completed a questionnaire to determine their status as a varsity athlete at the university and the amount of time they dedicated every week to physical activity. The relationship between proprioceptive and interoceptive accuracy scores was analyzed across the entire sample and compared across athletes and non-athletes. Despite differences in physical activity between athletes and non-athletes (p = 0.04, d = 0.87), results revealed no significant relationships between proprioceptive accuracy and interoceptive accuracy in the entire sample (r(22) = 0.16, p = 0.47), in athletes (r(9) = -0.31, p = 0.36), or in non-athletes (r(11) = 0.53, p = 0.06). Also, this relationship did not significantly differ between athletes and non-athletes (q = 0.91, p = 0.06), despite a large effect size. Overall, there were no significant relationships between athlete status, proprioceptive accuracy, and interoceptive accuracy.

Study on Downhole Geomagnetic Suitability Problems Based on Improved Back Propagation Neural Network

The analysis of geomagnetic suitability is the basis and premise of geomagnetic matching navigation and positioning. A geomagnetic suitability evaluation model using mixed sampling and an improved back propagation neural network (BPNN) based on the gray wolf optimization (GWO) algorithm by incorporating the dimension learning-based hunting (DLH) search strategy algorithm was proposed in this paper to accurately assess the geomagnetic suitability. Compared with the traditional geomagnetic suitability evaluation model, its generalization ability and accuracy were better improved. Firstly, the key indicators and matching labels used for geomagnetic suitability evaluation were analyzed, and an evaluation system was established. Then, a mixed sampling method based on the synthetic minority over-sampling technique (SMOTE) and Tomek Links was employed to extend the original dataset and construct a new dataset. Next, the dataset was divided into a training set and a test set, according to 7:3. The geomagnetic standard deviation, kurtosis coefficient, skewness coefficient, geomagnetic information entropy, geomagnetic roughness, variance of geomagnetic roughness, and correlation coefficient were used as input indicators and put into the DLH-GWO-BPNN model for model training with matching labels as output. Accuracy, recall, the ROC curve, and the AUC value were taken as evaluation indexes. Finally, PSO (Particle Swarm Optimization)-BPNN, WOA (Whale Optimization Algorithm)-BPNN, GA (Genetic Algorithm)-BPNN, and GWO-BPNN algorithms were selected as compared methods to verify the predictable ability of the DLH-GWO-BPNN. The accuracy ranking of the five models on the test set was as follows: PSO-BPNN (80.95 %) = WOA-BPNN (80.95%) &lt; GA-BPNN (85.71%) = GWO-BPNN (85.71%) &lt; DLH-GWO-BPNN (95.24%). The results indicate that the DLH-GWO-BPNN model can be used as a reliable method for underground geomagnetic suitability research, which can be applied to the research of geomagnetic matching navigation.