Electrical impedance tomography (EIT) has attracted increasing attention in medical imaging due to its noninvasive, portable, and real-time imaging capabilities. However, conventional EIT reconstruction methods still struggle to achieve high spatial resolution while maintaining physical consistency, particularly in pulmonary applications. In this work, we propose a novel Physics-Guided Cooperative Training framework based on PINN (PGCT-PINN) that enhances both resolution and consistency in lung EIT imaging. The proposed approach integrates the EIT forward model into a dual-branch architecture, comprising a voltage distribution prediction network (V-Net) and a conductivity reconstruction network ( σ - Net), while introducing cooperative training and physics-guided loss functions to leverage both data-driven features and physical constraints. High-resolution 3D thorax-lung simulation datasets are generated from CT priors via finite element modeling to enable robust training under both data-sufficient and data-limited conditions. Extensive simulation and phantom experiments demonstrate that PGCT-PINN achieves higher reconstruction accuracy, clearer structural boundaries, and stronger generalization capability, outperforming representative data-driven baselines such as Pix2Pix and U-Net, when training data are limited. Furthermore, a lightweight version of PGCT-PINN is deployed on a Raspberry Pi-based embedded platform, achieving real-time lung imaging at 16 frames per second, thus offering a practical solution for bedside pulmonary monitoring and early disease detection.

Light Detection and Ranging (LiDAR) encounters three-dimensional imaging degradation and noise generation in adverse weathers. To investigate its mechanism, an echo waveform model of LiDAR is developed in rain and fog, based on estimation of photons-return reception probability by semi-analytic Monte Carlo method. By innovatively combining of phase function for scattering and reflection, the proposed model can effectively depict both clutter and target waveforms and accurately estimated waveform parameters. Massive experiments are conducted to validate the method using significance testing (T-test), involving clutter and target waveforms in three weathers (i.e. heavy rain, heavy fog and light fog) and four key parameters (i.e. peak value, peak time, pulse width and skewness). Results show highly consistency between simulated and measured waveforms, as their statistical significances are well above typical significance value. The proposed model can further assist noise suppression, performance evaluation, and system design of LiDAR in complex weathers.

High-accuracy point cloud registration is an optimization solution by minimizing the alignment error of feature-metric correspondences. A novel compatibility distance ranking for maximal clique (CDR-MAC) based registration method is proposed for fast and high-accuracy point cloud registration. Based on the graph theory, the correspondences are constructed into graphs of multiple nodes. According to the calculation of edge between two nodes (i.e. compatible correspondences), a fast and valid maximal clique searching method is advanced for accurate transformation estimation, mainly by reserving small percentage of the correspondences after compatibility distance ranking. Extensive experiments on open-source 3DMatch dataset demonstrate that CDR-MAC method achieves the registration of rotation error of 1.50° and translation error of 0.130 m. The runtime of CDR-MAC is less than 2.97 s with an execution time speed-up of one order of magnitude. The method proposed in this work can be applied in the fields of mapping, target recognition and reconstruction.

The generation of dense and accurate point clouds from airborne light detection and ranging (LiDAR) waveform data is crucial to forest inventory. This work proposes a deconvolution method with: 1) an automatic stopping criterion to differentiate near-adjacent targets and 2) an iterative false subwaveform removal algorithm to remove outliers caused by noise. Synthetic waveforms with different overlap rates were processed using the proposed method, the Gaussian decomposition (GD) method, and the Richardson Lucy (RL) deconvolution method. Results showed that: 1) the number of subwaveforms detected by the proposed method is 9% higher than that of the RL and 20% higher than that of the GD when the overlap rates are larger than 0.6 and 2) the proposed method is of the smallest ground and peak distance errors. Results of the indoor experiment also show that the proposed method is superior in finding near targets meanwhile leading to small ground and peak distance error. Furthermore, the proposed method was tested by airborne waveforms from the Dagujia forest farm. The point cloud density acquired by the proposed method is 3% and 35% higher than that by the RL and GD method. Fewer outliers are produced by the proposed method. The number of individual trees extracted from the proposed point clouds is 22%, 51%, and 57% greater than those extracted from the RL, the GD, and the reference point clouds using the canopy height model-based method. Best individual tree extraction result is produced by the proposed method, especially for an area with small trees.

This paper proposes a low-cost, wearable gesture recognition system based on the two-terminal electrical impedance tomography (EIT) technique. The system includes a wearable EIT sensor of eight electrodes, a hardware device, and gesture recognition software running on a PC. Nine different gestures can be stably identified from the measured impedance changes through machine learning algorithms. Experimental results show that the Quadric Discriminator algorithm has the highest recognition rate of 98.49% for the filtered validation set. Besides, the recognition results in the two-terminal mode and transformed four-terminal mode are compared by applying a two-to-four-terminal mapping to the two-terminal EIT system, and the recognition rate decreases with the most classification models in the latter mode. Thus, it is supposed that contact impedance plays an important role in gesture recognition. By analyzing the data characteristics with variance inflation factor (VIF) test and principal component analysis (PCA), the supposition is explained and verified, proving the merit of a two-terminal EIT system in gesture recognition.

In precision farming, the separation of crops from soil is crucial for monitoring growth and fertilization. In this letter, a novel method is proposed for the accurate detection of corn seedlings from cropland by combining terrestrial laser scanning (TLS) and camera data. First, a piecewise linear interpolation method was used to eliminate the effect of distance on the TLS intensity data for more accurate intensity features of scanned targets. Second, the point cloud and camera data were registered to obtain the true color of each point in the point cloud. Third, we used a random forest algorithm to separate corn seedlings from soil by combining the geometric features from the TLS data with the radiometric features including the corrected intensity and RGB values derived from the TLS and camera data. To evaluate the proposed method, a case study was conducted by using a commercial TLS sensor with an embedded camera. The results demonstrated that corn seedlings can be separated from soil with an accuracy of 98.8% by using both the geometric and radiometric features, which is significantly higher than that by using any one of the two kinds of features.

Point clouds registration is an important research topic in the field of data fusion from camera and light detection and ranging (LiDAR). In this letter, a new registration method, fast multiscale registration (FMSR), takes the scale factor into account and is proposed for the registration of two point clouds obtained from camera and LiDAR. An adaptive-scale keypoint quality algorithm was used to detect and match keypoints, which were input to the coarse registration process to improve the coarse registration accuracy. A new heuristic criterion was also proposed for fine registration, which avoids falling into the local minima. Furthermore, to increase efficiency of fine registration, the k-nearest neighbors algorithm was selected to directly search the optimal matching from the raw point clouds without triangulating point clouds into mesh. The FMSR method is highly precise, insensitive to outliers, and relatively efficient. Experimental results showed that the root-mean-square error of the registration was approximately 0.2 m when the size of the object was about 20.3 m × 7.85 m × 26.56 m, the total number of matched points was 12 789, and the execution time was approximately 2.1 s, indicating that the proposed method resulted in improved accuracy and efficiency of registration.

The increased requirement for visible light/infrared dual band imaging systems with compact structure, suitable field of view (FOV), high resolution coverage and without significantly complicates the optical design is increasing. Such systems are particularly desirable for unmanned aerial vehicle (UAV) remote sensing applications where increased levels of target reconnaissance, surveillance, precision target location and designation are required in cost-effective and light weight systems. Therefore, in this paper, a high resolution Electro-optical/infrared (EO/IR) sensor to simultaneously operate in the visible band and short wave infrared band (SWIR) band using ZEMAX software are presented. It has a common aperture using common refractive system and the spectrum is separated using a dichroic beam splitter. A long wave infrared (LWIR) imaging system based on uncooled microbolometer focal plane array (FPA) detector with pixel size equal to 17μm is presented. The results of the optical systems design indicated that our proposed approaches meet the technical performance requirements for high image performance and compact size. This system can be applied in UAV remote sensing applications.

This letter presents an on-the-fly method for extracting polyhedral buildings from airborne light detection and ranging (LiDAR) data. By using the gridding method, the planimetric position and elevation of laser footprints (normally treated as points) in the obtained scan line are mapped into a data sequence. Then, discrete stationary wavelet transform is applied to analyze the elevation variation in the sequence. Buildings in the scan line can be obtained from the detail wavelet coefficients of the sequence. Moreover, to improve precision of the extraction, the gradients of grid points in the geometric planes of building roofs along the direction of the scan line are calculated and remedied by using the corresponding gradients acquired from the adjacent scan lines. With the proposed on-the-fly method, polyhedral buildings in the scan area can be accurately extracted from laser points along the scan lines during the scanning process. The new method is validated by using a set of real airborne LiDAR data.

In this paper, slope estimation model and algorithms were advanced using the repeat ground track data collected from the Geoscience Laser Altimeter System (GLAS) on board the Ice, Cloud and land Elevation Satellite (ICESat). Innovative steps of the algorithm are to prove and confirm the reference starting point (RSF) on reference track (RT) and the corresponding starting point (CSF) on other repeat corresponding track (CT). The slopes model and algorithms were validated by the ground track 100 data which were collected from ICESat/GLAS passing over the Beijing area during 2003-2009. The slopes drawing on the geographic map were coincidence with the geography facts compared with image obtained from Google map.

As a new case of Light Detection and Ranging (LiDAR), full-waveform LiDAR records the complete waveform of backscattered echo of targets in certain time interval using high-speed data acquisition device. Since the full-waveform signal is generally short in length and badly contaminated by noise, it is rather difficult to find a method suitable for the signal filtering. In this paper, the Empirical Mode Decomposition (EMD) was extended to the filtering of full-waveform LiDAR signal. Aiming at simulation signal, the filtering results of EMD-based filtering method were respectively compared with those deduced from Low-pass filter, Wiener filter and Gaussian smoothing. The filtering results show that the Signal to Noise Improvement Ratio (SNIR) of EMD-based filtering method is biggest in all compared filtering methods. Residual Sum of Squares (RSS) of EMD-based filtering method is just bigger than Wiener filter. Meanwhile, the processing results of different filtering methods were fitting with Gaussian function using Levenberg-Marquardt (LM) method. Based on the compare of fitting parameters accuracy of signal filtered by different filtering methods, EMD-based method is more suitable for the preprocessing of Gaussian fitting. At the last, some typical Geoscience Laser Altimeter System (GLAS) data were filtered and fitted using EMD-based filtering method and Levenberg-Marquardt fitting method. The experimental results suggest that the EMD-based filtering method has well filtering result.

In this paper, a novel physical model and its mathematical expressions were used to estimate the average slope of land by using LVIS data, which was collected in mountain area, California, USA. The waveform data obtained from LVIS were used to validate the local slope within footprint. It was compared with the other two methods which were based on the mean elevation of land and the forest canopy height. For eliminating the variance between the method proposed in this paper and two compared method, the mean elevation of land was used to modify the result estimated by the method advanced above. Calculation results indicate that the method proposed in this paper can make a more accurate and average estimation for the local slope of land and slope estimation was not influenced by the forest canopy height too much.

In this paper, a novel physical model and its mathematical expressions were advanced to estimate the average slope of land by using LVIS data, which was collected in mountain area, California, USA. The proposed expressions were related not only to the laser scan angle and the laser divergence, but also to the pulse delay and the pulse width. In order to calculate the time width of the echo waveform and the mean pulse delay, a new curve-fitting method was proposed to fit the echo waveforms using a non-Gaussian mathematical function named as ‘Extreme function’. Different from the slope estimation from the DEM, the method proposed in this paper can make a more average estimation for the local slope within laser footprint. It can be used to calculate the average slope of land and estimate the roughness roughly.