On-board Detection Research Articles

Automatic Detection of Quasi-Periodic Emissions from Satellite Observations by Using DETR Method

The ionospheric quasi-periodic wave is a type of typical and common electromagnetic wave phenomenon occurring in extremely low-frequency (ELF) and very low-frequency ranges (VLF). These emissions propagate in a distinct whistler-wave mode, with varying periodic modulations of the wave intensity over time scales from several seconds to a few minutes. We developed an automatic detection model for the QP waves in the ELF band recorded by the China Seismo-Electromagnetic Satellite. Based on the 827 QP wave events, which were collected through visual screening from the electromagnetic field observations, an automatic detection model based on the Transformer architecture was built. This model, comprising 34.27 million parameters, was trained and evaluated. It achieved mean average precision of 92.3% on the validation dataset, operating at a frame rate of 39.3 frames per second. Notably, after incorporating the proton cyclotron frequency constraint, the model displayed promising performance. Its lightweight design facilitates easy deployment on satellite equipment, significantly enhancing the feasibility of on-board detection.

A photometric-turbidimetric fusion sensor for measuring the water content in diesel with sensitivity increase by the extended infrared path capillary

Petroleum-based diesel is the main energy supply for heavy-duty vehicles and industrial equipment. Excessive water in the diesel would affect engine performance directly and even lead to serious accidents. In this paper, a new on-board compact sensor for detecting water content in diesel is proposed, which is composed of LED, photodetector, special metal waveguide capillary and ultrasonic cavity. Especially, the metal waveguide capillary with extended optical path can significantly improves the measurement accuracy meanwhile decrease the size. The small ultrasonic oscillation cavity is connected with the metal waveguide capillary to ensure the saturated water droplets to be evenly mixed with petroleum-based diesel before being detected. Then, the photometric - turbidimetric fusion method is proposed to measure the total water contents in different states (i.e., dissolved water, emulsified water and free water) in the diesel. The designed metal waveguide capillary with extended optical path improves sensor accuracy by the way of multiple reflections of light. In addition, the sensor is installed in the line between the fuel tank and the engine, then, the actual water pumped into the engine can be accurately measured. The experimental results show that the root mean square of the prediction error (RMSEP) is only 7.34 ppm, which is much lower than the optical sensor based on non-dispersive infrared spectroscopy of 18.6 ppm. The sensor fabricated in this paper finds application in detecting water content in diesel, which has the advantages of high sensitivity, small volume, and are suitable for on-board detection.

SCCMDet: Adaptive Sparse Convolutional Networks Based on Class Maps for Real-Time Onboard Detection in Unmanned Aerial Vehicle Remote Sensing Images

Onboard, real-time object detection in unmaned aerial vehicle remote sensing (UAV-RS) has always been a prominent challenge due to the higher image resolution required and the limited computing resources available. Due to the trade-off between accuracy and efficiency, the advantages of UAV-RS are difficult to fully exploit. Current sparse-convolution-based detectors only convolve some of the meaningful features in order to accelerate the inference speed. However, the best approach to the selection of meaningful features, which ultimately determines the performance, is an open question. This study proposes the use of adaptive sparse convolutional networks based on class maps for real-time onboard detection in UAV-RS images (SCCMDet) to solve this problem. For data pre-processing, SCCMDet obtains the real class maps as labels from the ground truth to supervise the feature selection process. In addition, a generate class map network (GCMN), equipped with a newly designed loss function, identifies the importance of features to generate a binary class map which filters the image for its more meaningful sparse features. Comparative experiments were conducted on the VisDrone dataset, and the experimental results show that our method accelerates YOLOv8 by 41.94% at most and increases the performance by 2.52%. Moreover, ablation experiments demonstrate the effectiveness of the proposed model.

Rapid detection of microalgae cells based on upconversion nanoprobes.

The quantification of microalgae cells is crucial for the treatment of ships' ballast water. However, achieving rapid detection of microalgae cells remains a substantial challenge. Here, we develop a new method for rapid and effective detection of microalgae concentration by utilizing upconversion nanoprobes (UCNPs) of NaYF4:Er3+,Tm3+. Three ligands, carboxylated methoxypolyethylene glycols with 5000 and 2000 molecular weights (mPEG-COOH-5, mPEG-COOH-2) and D-gluconic acid sodium salt (DGAS), were used to convert hydrophobic UCNPs into a hydrophilic state through modification. The results show that the mPEG-COOH-5 modified UCNPs present the highest stability in an aqueous solution. Fourier Transform Infrared Spectroscopy (FTIR) measurements reveal the presence of a significant number of -COOH functional groups on UCNPs after the mPEG-COOH-5 modification. These -COOH groups enhance the hydrophilicity and biocompatibility of UCNPs. The soluble UCNPs were directly mixed with microalgae, and the upconversion luminescence (UCL) spectra of the UCNPs were recorded immediately after thorough shaking. This greatly reduces the measurement time and could realize rapid onboard detection. In this sensing procedure, the UCNPs with red UCL functioned as energy donors, while microalgae with red absorption served as an energy acceptor. The UCL gradually diminishes with an increase in microalgae concentration based on the inner filter effect, thus establishing a relationship between UCL and microalgae concentration. The accuracy of the detection is further validated through the traditional microscope counting method. These findings pave the way for a novel rapid strategy to assess microalgae concentration using UCNPs.

Integration of complementary split-ring resonators into digital microfluidics for manipulation and direct sensing of droplet composition.

This paper demonstrates the integration of complementary split-ring resonators (CSSRs) with digital microfluidics (DMF) sample manipulation for passive, on-chip radio-frequency (RF) sensing. Integration is accomplished by having the DMF and the RF-sensing components share the same ground plane: by designing the RF-resonant openings directly into the ground plane of a DMF device, both droplet motion and sensing are achieved, adding a new on-board detection mode for use in DMF. The system was modelled to determine basic features and to balance various factors that need to be optimized to maintain both functionalities (DMF-enabled droplet movement and RF detection) on the same chip. Simulated and experimental results show good agreement. Using a portable measurement setup, the integrated CSSR sensor was used to effectively identify a series of DMF-generated drops of ethanol-water mixtures of different compositions by measuring the resonant frequency of the CSSR. In addition, we show that a binary solvent system (ethanol/water mixtures) results in consistent changes in the measured spectrum in response to changes in concentration, indicating that the sensor can distinguish not only between pure solvents from each other, but also between mixtures of varied compositions. We anticipate that this system can be refined further to enable additional applications and detection modes for DMF systems and other portable sensing platforms alike. This proof-of-principle study demonstrates that the integrated DMF-CSSR sensor provides a new platform for monitoring and characterization of liquids with high sensitivity and low consumption of materials, and opens the way for new and exciting applications of RF sensing in microfluidics.

Use of a Residual Neural Network to Demonstrate Feasibility of Ship Detection Based on Synthetic Aperture Radar Raw Data

Synthetic Aperture Radar (SAR) is a well-established 2D imaging technique employed as a consolidated practice in several oil spill monitoring services. In this scenario, onboard detection undoubtedly represents an interesting solution to reduce the latency of these services, also enabling transmission to the ground segment of alert signals with a notable reduction in the required downlink bandwidth. However, the reduced computational capabilities available onboard require alternative approaches with respect to the standard processing flows. In this work, we propose a feasibility study of oil spill detection applied directly to raw data, which is a solution not sufficiently addressed in the literature that has the advantage of not requiring the execution of the focusing step. The study is concentrated only on the accuracy of detection, while computational cost analysis is not within the scope of this work. More specifically, we propose a complete framework based on the use of a Residual Neural Network (ResNet), including a simple and automatic simulation method for generating the training data set. The final tests with ERS real data demonstrate the feasibility of the proposed approach showing that the trained ResNet correctly detects ships with a Signal-to-Clutter Ratio (SCR) > 10.3 dB.

SEADETECT: developing an automated detection system to reduce whale-vessel collision risk

With the continuous intensification of marine traffic worldwide, whale-vessel collisions at sea (or “ship strikes”) have become one of the primary causes of mortality for cetaceans and a widely recognised cause of concern for human safety and economic losses. The Mediterranean Sea is a global hotspot for whale-vessel collisions, with one of the highest rates involving large cetaceans, especially the endangered fin whales (Balaenoptera physalus) and sperm whales (Physeter macrocephalus). Evidence indicates that both species are experiencing higher chances of a fatal collision than what predictions have estimated so far, with ship strikes being the main human-induced threat in the area. Regional and international organisations have stressed the need to address the issue by investigating the projected impacts of ship strikes on whale populations and by identifying possible mitigation measures to reduce chances of collision. Amongst the most popular and feasible options, there is the improvement of animal detection during navigation. Here, we present SEADETECT, a LIFE project that aims at developing an automated detection system to reduce vessel collision risk with marine mammals and unidentified floating objects (UFOs), combining state-of-the-art and novel technologies with existing approaches in the study of large whale ecology. This detection system consists of three elements; an automated onboard detection system composed of several sensors, a real-time passive acoustic monitoring (PAM) network at sea and a real-time detection-sharing and alert system (REPCET®). In this paper, we propose the development of a mitigation measure framework tailored for the issue of collision with fin and sperm whales in the north-western Mediterranean Sea, but that has the transferability features necessary for its application in other high-risk areas for ship strikes worldwide.

Gaia data processing

Aims. Our goal is to describe two potential options for the Source Environment Analysis pipeline (SEAPipe) for the Gaia mission. This pipeline will enable the discovery of sources that are new to Gaia, in the sense that they were not found by the on-board detection algorithm. These additional sources (secondaries) are discoverable in the vicinity of those Gaia sources (primaries) that were found by the on-board detection. Methods. The main algorithmic steps required are described: the two dimensional image reconstruction of one dimensional transit data; the analysis of these images to find the additional sources present, and the determination of the mean positions; proper motions, parallaxes, and brightness of these sources. Additionally, the Monte Carlo simulations used to characterise the performance of the pipelines are described. Results. The performance of the two options for SEAPipe, the vanilla and image-subtraction versions, are compared. Their selection functions were computed in terms of the magnitude of the secondary sources and their angular separations from their corresponding primary source. The completeness and purity of the resultant catalogue of secondary sources as found by each of the pipelines, given the expected magnitude distribution of the primary sources and the magnitude and angular separation distributions of the secondary sources, is also presented. The image-subtraction pipeline is shown to outperform the vanilla pipeline.

Wheel Defect Detection Using a Hybrid Deep Learning Approach.

Defective wheels pose a significant challenge in railway transportation, impacting operational performance and safety. Excessive traction and braking forces give rise to deviations from the intended conical tread shape, resulting in amplified vibrations and noise. Moreover, these deviations contribute to the accelerated damage of track components. Detecting wheel defects at an early stage is crucial to ensure safe and comfortable operation, as well as to minimize maintenance costs. However, the presence of various vibrations, such as those induced by the track, traction motors, and other rolling stock subsystems, poses a significant challenge for onboard detection techniques. These vibrations create difficulties in accurately identifying wheel defects in real-time during operational activities, often resulting in false alarms. This research paper aims to address this issue by using a hybrid deep learning-based approach for the accurate detection of various types of wheel defects using accelerometer data. The proposed approach aims to enhance wheel defect detection accuracy while considering onboard techniques' cost-effectiveness and efficiency. A realistic simulation model of the railway wheelset is developed to generate a comprehensive dataset. To generate vibration data in various scenarios, the model is simulated for 20 s under different conditions, including one non-faulty scenario and six faulty scenarios. The simulations are conducted at different speeds and track conditions to capture a wide range of operating conditions. Within each simulation iteration, a total of 200,000 data points are generated, providing a comprehensive dataset for analysis and evaluation. The generated data are then utilized to train and evaluate a hybrid deep learning model, employing a multi-layer perceptron (MLP) as a feature extractor and multiple machine learning models (support vector machine, random forest, decision tree, and k-nearest neighbors) for performance comparison. The results demonstrate that the MLP-RF (multi-layer perceptron with random forest) model achieved an accuracy of 99%, while the MLP-DT (multi-layer perceptron with decision tree) model achieved an accuracy of 98%. These high accuracy values indicate the effectiveness of the models in accurately classifying and predicting the outcomes. The contributions of this research work include the development of a realistic simulation model, the evaluation of sensor layout effectiveness, and the application of deep learning techniques for improved wheel flat detections.

An anti-disturbance method for on-board detection of early wheel polygonal wear by weighted angle-synchronous moving average

Wheel polygonal wear (WPW) on-board detection based on axle-box acceleration (ABA) is a critical instrument for railway wheel maintenance. However, early WPW on-board detection under typical disturbances such as rail defects is still difficult to achieve. To face this challenge, this study proposes a weighted angle-synchronous moving average (WASMA) anti-disturbance filtering for WPW on-board detection. Starting with the gear mesh component, the accurate instantaneous phase of the wheelset is calculated using a phase demodulation and time–frequency ridges combination. The ABA is then filtered using WASMA to suppress typical disturbances while revealing the features of early WPW. Simulation and field tests were carried out to verify the proposed detection method. The results demonstrated that the proposed method effectively filters out typical disturbances, such as random track irregularity, sleeper passing impact, and structural resonance, reducing the misdiagnosis of early WPW.

On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs)

Object detection is a critical task that becomes difficult when dealing with onboard detection using aerial images and computer vision technique. The main challenges with aerial images are small target sizes, low resolution, occlusion, attitude, and scale variations, which affect the performance of many object detectors. The accuracy of the detection and the efficiency of the inference are always trade-offs. We modified the architecture of CenterNet and used different CNN-based backbones of ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, Res2Net50, Res2Net101, DLA-34, and hourglass14. A comparison of the modified CenterNet with nine CNN-based backbones is conducted and validated using three challenging datasets, i.e., VisDrone, Stanford Drone dataset (SSD), and AU-AIR. We also implemented well-known off-the-shelf object detectors, i.e., YoloV1 to YoloV7, SSD-MobileNet-V2, and Faster RCNN. The proposed approach and state-of-the-art object detectors are optimized and then implemented on cross-edge platforms, i.e., NVIDIA Jetson Xavier, NVIDIA Jetson Nano, and Neuro Compute Stick 2 (NCS2). A detailed comparison of performance between edge platforms is provided. Our modified CenterNet combination with hourglass as a backbone achieved 91.62%, 75.61%, and 34.82% mAP using the validation sets of AU-AIR, SSD, and VisDrone datasets, respectively. An FPS of 40.02 was achieved using the ResNet18 backbone. We also compared our approach with the latest cutting-edge research and found promising results for both discrete GPU and edge platforms.

UAVs Beneath the Surface: Cooperative Autonomy for Subterranean Search and Rescue in DARPA SubT

This paper presents a novel approach for autonomous cooperating UAVs in search and rescue operations in subterranean domains with complex topology. The proposed system was ranked second in the Virtual Track of the DARPA SubT Finals as part of the team CTU-CRAS-NORLAB. In contrast to the winning solution that was developed specifically for the Virtual Track, the proposed solution also proved to be a robust system for deployment onboard physical UAVs flying in the extremely harsh and confined environment of the real-world competition. The proposed approach enables fully autonomous and decentralized deployment of a UAV team with seamless simulation-toworld transfer, and proves its advantage over less mobile UGV teams in the flyable space of diverse environments. The main contributions of the paper are present in the mapping and navigation pipelines. The mapping approach employs novel map representations—SphereMap for efficient risk-aware long-distance planning, FacetMap for surface coverage, and the compressed topologicalvolumetric LTVMap for allowing multirobot cooperation under low-bandwidth communication. These representations are used in navigation together with novel methods for visibility-constrained informed search in a general 3D environment with no assumptions about the environment structure, while balancing deep exploration with sensor-coverage exploitation. The proposed solution also includes a visual-perception pipeline for on-board detection and localization of objects of interest in four RGB stream at 5 Hz each without a dedicated GPU. Apart from participation in the DARPA SubT, the performance of the UAV system is supported by extensive experimental verification in diverse environments with both qualitative and quantitative evaluation.

Onboard early detection and mitigation of lithium plating in fast-charging batteries

Fast-charging is considered as one of the most desired features needed for lithium-ion batteries to accelerate the mainstream adoption of electric vehicles. However, current battery charging protocols mainly consist of conservative rate steps to avoid potential hazardous lithium plating and its associated parasitic reactions. A highly sensitive onboard detection method could enable battery fast-charging without reaching the lithium plating regime. Here, we demonstrate a novel differential pressure sensing method to precisely detect the lithium plating event. By measuring the real-time change of cell pressure per unit of charge (dP/dQ) and comparing it with the threshold defined by the maximum of dP/dQ during lithium-ion intercalation into the negative electrode, the onset of lithium plating before its extensive growth can be detected with high precision. In addition, we show that by integrating this differential pressure sensing into the battery management system (BMS), a dynamic self-regulated charging protocol can be realized to effectively extinguish the lithium plating triggered by low temperature (0 °C) while the conventional static charging protocol leads to catastrophic lithium plating at the same condition. We propose that differential pressure sensing could serve as an early nondestructive diagnosis method to guide the development of fast-charging battery technologies.

ConstDet: Control Semantics-Based Detection for GPS Spoofing Attacks on UAVs

UAVs are widely used in agriculture, the military, and industry. However, it is easy to perform GPS spoofing attacks on UAVs, which can lead to catastrophic consequences. In this paper, we propose ConstDet, a control semantics-based detection approach for GPS spoofing attacks of UAVs using machine learning algorithms. Various real experiments are conducted to collect real flight data, on the basis of which ConstDet is designed as a practical detection framework. To train models for the detection of GPS spoofing attacks, specified flight data types are selected as features based on the control semantics, including the altitude control process and the horizontal position control process, since these data are able to represent the dynamic flight and control processes. Multiple machine learning algorithms are used to train and generate the best classifier for GPS spoofing attacks. ConstDet is further implemented and deployed on a real UAV to support onboard detection. Experiments and evaluations validate that ConstDet can effectively detect GPS spoofing attacks and the detection rate can reach 97.70%. The experimental comparison demonstrates that ConstDet has better performance than existing detection approaches.

Deep neural network‐driven in‐situ detection and quantification of lithium plating on anodes in commercial lithium‐ion batteries

AbstractLithium plating seriously threatens the life of lithium‐ion batteries at low temperatures charging conditions, but the onboard detection and quantification of lithium plating are severely hampered by the limited available signals and volatile operating conditions in real scenarios. Herein, we propose a detection method to predict the occurrence and quantification of lithium plating under uncertain conditions by only using constant‐current curves during charging based on deep learning. A deep neural network (DNN) is developed to extract data‐driven features induced by lithium plating from the charge curves, avoiding the challenge of manual feature selection. Only using the most common voltage and current signals as inputs, the network exhibits superior adaptability and accuracy. The detection accuracy of the proposed method is 98.64%, while the quantity of the lithium plating can be accurately predicted with a root‐mean‐square error <4.1712 mg. Moreover, the generalization ability of the proposed method is verified by its reliable detection accuracy under conditions that are not used in the training dataset. The detection accuracy is 92.39% for brand new charging conditions and 95.53% for brand new aging states. This method shortens the detection time that currently takes more than several hours (the widely used differential curve analysis) to milliseconds and eliminates the need for a rigorous testing environment, showing great potential for onboard application in future battery management systems.image

Automated detection of gunshots in tropical forests using convolutional neural networks

Unsustainable hunting is one of the leading drivers of global biodiversity loss, yet very few direct measures exist due to the difficulty in monitoring this cryptic activity. Where guns are commonly used for hunting, such as in the tropical forests of the Americas and Africa, acoustic detection can potentially provide a solution to this monitoring challenge. The emergence of low cost autonomous recording units (ARUs) brings into reach the ability to monitor hunting pressure over wide spatial and temporal scales. However, ARUs produce immense amounts of data, and long term and large-scale monitoring is not possible without efficient automated sound classification techniques. We tested the effectiveness of a sequential two-stage detection pipeline for detecting gunshots from acoustic data collected in the tropical forests of Belize. The pipeline involved an on-board detection algorithm which was developed and tested in a prior study, followed by a spectrogram based convolutional neural network (CNN), which was developed in this manuscript. As gunshots are rare events, we focussed on developing a classification pipeline that maximises recall at the cost of increased false positives, with the aim of using the classifier to assist human annotation of files. We trained the CNN on annotated data collected across two study sites in Belize, comprising 597 gunshots and 28,195 background sounds. Predictions from the annotated validation dataset comprising 150 gunshots and 7044 background sounds collected from the same sites yielded a recall of 0.95 and precision of 0.85. The combined recall of the two-step pipeline was estimated at 0.80. We subsequently applied the CNN to an un-annotated dataset of over 160,000 files collected in a spatially distinct study site to test for generalisability and precision under a more realistic monitoring scenario. Our model was able to generalise to this dataset, and classified gunshots with 0.57 precision and estimated 80% recall, producing a substantially more manageable dataset for human verification. Using a classifier-guided listening approach such as ours can make wide scale monitoring of threats such as hunting a feasible option for conservation management.

Investigation and On-Board Detection of Gate-Open Failure in SiC MOSFETs

Gate-open failures in power semiconductors occur when the gate-bond wire cracks or lifts-off leading to loss of gate control. In molded discrete devices, this failure mode may occur intermittently making it very challenging to analyze and detect. In this paper, intermittent gate-open failures are comprehensively investigated in the context of discrete SiC MOSFETs. First, the MOSFETs behavior under various possible gate-open failure scenarios is analyzed in detail through simulations. Several SiC MOSFETs are aged on a DC power cycling setup and gate-open failure mechanism is verified through systematic multi-step failure analysis which includes on-board characterization, non-destructive C-SAM analysis, decapsulation and optical inspection followed by SEM analysis of the failed devices. To understand the potential mechanism behind gateopen failure in SiC MOSFETs, thermo-mechanical FEA analysis is performed on a high-fidelity model which shows interfacial shear stress at gate-bond. Further, a robust on-board technique for reliable cycle-by-cycle detection of gate-open faults is proposed. The proposed technique is experimentally verified for all possible fault scenarios and shown to detect faults in as low as 150ns. It is shown that compared to traditional DESAT protection scheme, the proposed mechanism can prevent potential shoot-through events that may be caused by gate-open failure.

Unsupervised detection of Saturn magnetic field boundary crossings from plasma spectrometer data

We investigate unsupervised time-series analysis techniques for real-time identification of magnetic field boundaries in data collected by the Cassini Plasma Spectrometer (CAPS) on the Cassini mission to Saturn. Previous research has sought to identify these boundaries either in a supervised setting, where a subset of labeled boundary crossing events are available to the algorithms, or using magnetometer data in which these boundaries are significantly clearer. We find that Saturn bow shock transitions can be reliably detected by the methods we consider, while magnetopause transitions are harder to identify across different years, indicating that generalization of algorithm parameters remains a challenge. We identify the Plasma Instrument for Magnetic Sounding (PIMS) on the upcoming Europa Clipper Mission as a promising beneficiary of the use of similar processing methods. By performing onboard detection of events such as Jovian plasma sheet crossings, Europa ionosphere boundaries, and water vapor plumes, Europa Clipper could benefit from in-situ adaptation to increase science return and data quality, such as by changing the instrument’s observing mode depending on the spacecraft’s current environment.

Fast onboard detection of ice drift changes under stationkeeping in ice

Early detection of ice drift changes is essential for safe stationkeeping in ice-infested waters. This paper presents a conceptual system allowing onboard detection of ice drift changes in managed and unmanaged ice using distributed accelerometers installed in the hull of a stationkeeping vessel. With methods from statistical signal processing, specifically probability of exceedance analysis, a relation between the measured signals' entropy and the location along the hull where the ice action occurs is found. With field data from a stationkeeping operation in the Bay of Bothnia this relation is validated. The main contribution of this paper is a concept for ice drift estimation which only requires distributed vibration measurements along the hull and the ship's heading information.

Effectiveness of surface-based detection methods for vessel strike mitigation of North Atlantic right whales

Increasing commercial and recreational use of the world’s oceans has led to growing concerns about vessel and marine mammal encounters. For endangered species such as the North Atlantic right whale Eubalaena glacialis, reducing the number of vessel strikes is key to improving their protection. In this study, we developed an agent-based model to assess the efficacy of thermal imaging systems as a surface-based whale detection method for vessel strike mitigation. We found that the detection range of such systems is the determining factor for their efficacy and needs to be chosen according to vessel characteristics, such as speed and maneuverability. Furthermore, we found that combining large-scale (e.g. protected zones) and small-scale (e.g. on-board detection systems) mitigation strategies increases protection. Finally, technological improvements are needed to achieve reliable detection ranges beyond what is currently possible so that fast and poorly maneuverable vessels such as ultra-large container ships could benefit from on-board detection systems.