This letter revisits the differential electric-field-sensitive field-effect transistor (DeFET), a 0.18-$\mu$m CMOS sensor that converts the potential difference induced across adjacent floating gates into a complementary current difference. The revision consolidates the device, interface, noise, and mismatch models and reexamines archived electrical and microfluidic measurements. The measured air transfer characteristic is linear from $-3$ to $+3$ V/$\mu$m ($R^{2}=0.997$), with 51.7mV/(V/$\mu$m) sensitivity and 25-$\mu$ V null offset. The prototype consumed 1.23mW, showed 17/15-ns rise/fall times, and exhibited 11- and 12.5-MHz bandwidths in air and silicone rubber, respectively. A spectrum-analyzer floor of approximately $-150$ to $-144$ dBm/Hz corresponds conditionally to 0.14–0.27 $\mu$V/$\mu$m/$\sqrt{\text{Hz}}$ equivalent field noise for a 50-$\Omega$ termination. Separate microfluidic measurements in deionized water (1.3–1.9 $\mu$S/cm) detected field changes produced by 8.9- and 20.9-$\mu$m polystyrene particles; at 20 MHz the reported dry/wet output amplitudes were 830/844 mV. These results demonstrate fast local field sensing and medium/particle contrast, while operation in physiological buffers and with living cells remains to be validated.
The increasing prevalence of chronic cardiopulmonary diseases and the growing demand for remote healthcare solutions highlight the need for continuous and reliable physiological monitoring. This work proposes a flexible graphene-based multimodal sensor for simultaneous cardiorespiratory monitoring. Owing to the intrinsic properties of graphene-based nanocomposites, including high electrical conductivity, mechanical strength, and flexibility, the proposed sensor enables a direct multimodal approach using a reduced sensing architecture. Unlike approaches that require a dedicated respiratory sensor or additional respiratory sensing modality, this sensor acquires electrocardiogram (ECG) signals while extracting respiratory activity from variations at the electrode–skin interface from the same flexible electrode platform. This enables simultaneous monitoring of cardiac and respiratory activity while reducing sensing hardware complexity. Respiratory signals are validated against a reference force/pressure sensor, referred to as pressure-derived respiration (PDR). The proposed method is evaluated in terms of waveform similarity, frequency domain characteristics, temporal alignment, and respiratory rate (RR) agreement. The results demonstrate the accuracy and feasibility of the proposed approach in this proof-of-concept study, highlighting its potential for continuous and unobtrusive respiration monitoring in wearable healthcare applications.
Registering billion-scale LiDAR point clouds is a critical challenge in remote sensing and engineering. Standard in-core algorithms face prohibitive memory bottlenecks, while emerging deep learning methods are constrained by GPU memory limits. In this paper, we propose OOC-HICP, an Out-of-Core Hierarchical ICP framework designed to register massive COPC-formatted LiDAR datasets with CPU-only processing and bounded point-payload residency. The framework leverages the Cloud Optimized Point Cloud (COPC) indexing standard to enable a memory-efficient coarse-to-fine registration process. Specifically, we introduce a dynamic overlap querying mechanism that loads only spatially overlapping nodes from disk, combined with a budgeted active-set selection strategy. This approach decouples registration-stage memory consumption from the total data volume by refining one scene-level transform on metadata-guided active sets rather than materializing the full scene. A same-process run on a multi-temporal offshore platform dataset containing over 4.29 billion points completes Phase I and L7/L8/L9 refinement in 18.23 min with 7.85 GB registration-stage peak RSS. On a controlled 160M-point pair, OOC-HICP uses 2.37 GB registration-stage peak RSS and yields 0.16 cm known-transform translation error; evaluation on the same fixed 80M+80 M point files gives 69.65% Fitness and 1.06 cm inlier RMSE. A global-uniform baseline and three public WHU-TLS pairs provide additional controlled validation.
This letter presents a MEMS-based multi-turn inductive loop sensor (mILS) for detecting the transient magnetic field generated by partial discharge (PD) current pulses. The proposed planar sensor, operating below its self-resonant frequency (SRF), functions as a passive inductive magnetic-field probe with broadband HF/VHF response. Fabricated within a 2.44 × 2.44 mm² footprint, the sensor exhibits an intrinsic SRF of 320 MHz after de-embedding. Surface PD experiments demonstrate direct pulse detection without external amplification, producing an output of approximately 8 mV at a distance of 5 cm. The compact, passive, and low-cost design is well suited for integrated PD monitoring and smart insulation diagnostics.
This letter presents a 360-degree thermal imaging system for single-pass edge perception. VT360 uses six FLIR Lepton 3.5 modules arranged at 60$^\circ$ intervals and introduces border-local gain compensation followed by multi-band Laplacian pyramid blending (MBLPB) as its core panorama-construction method. The proposed method corrects intensity transitions near sensor boundaries and forms a unified thermal input for single-pass YOLO detection. Across ten YOLO variants, it outperforms split-stream detection on both validation metrics, with mean absolute gains of 9.45 percentage points in mAP@50 and 5.40 points in mAP@50–95. Relative to a hard-concatenation baseline, it increases mean mAP@50 by 1.19 points, with a 2.46-point reduction in mAP@50–95. Run-time inference evaluations are performed on Raspberry Pi, Rock Pi, and NVIDIA Jetson platforms to characterize the original single-pass detection pipeline and proposed methodology.
