1 December 2023 in IEEE Transactions on Instrumentation and Measurement
Capturing the motion of humans without any markers in a moving scene is a challenging issue. Typical methods, such as those based on fixed cameras or wearable sensors, have limits on the measurement range and the number of targets. Therefore, we propose a markerless motion capture technique to estimate the 3D pose of humans through a multi-UAV system. By utilizing the advantage of the large observation baselines between UAVs in the system, an algorithm to estimate the pose of airborne cameras and establish the sparse points cloud of background in overlapping view simultaneously was presented. A 3σ rule based on reprojection error is employed to eliminate the incorrect conjugate keypoint pairs, enhancing the precision of pose estimation for airborne cameras. To verify the accuracy of the developed technique, an indoor test was conducted to compare it with a commercial capture system, the results show that the average measurement error of the 3D joints of humans is less than 30mm. Furthermore, we demonstrate the feasibility of our approach in outdoor scenes, including multi-person scenarios and moving scenarios.
26 September 2023 in IEEE Sensors Journal
Rotary encoders are commonly used in industry for high precision motor control. This paper proposes a polar transformation of the digital image grating (DIG) method to realize a simple, accurate and low-cost rotary encoder. It measures angles with a rather small camera view to get high accuracy angle results and uses DIG strategy to realize full range measurement. The DIG method estimates the rough angle position by code mark recognition and obtains the accurate angle position by precise image matching. First, a circle image pattern is designed and is put on the top of the rotator. Second, images of the circle image pattern are grabbed and transformed to polar coordinate system with polar transformation method. Last, the angle positions are calculated from the transformed images based on the DIG method. Simulations show that the proposed method is capable of measuring rotary angles with MSER below 1 arcsec. In experiments, the angle result precisions are about 1.6 arcsecs (MSER) with a rotary diameter of 60 mm, and the accuracy is 10 times higher than the existing monocular methods.
8 September 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Multi-frame infrared small target detection (IRSTD) plays a crucial role in low-altitude and maritime surveillance. The hybrid architecture combining CNNs and Transformers shows great promise for enhancing multi-frame IRSTD performance. In the paper, we propose LVNet, a simple yet powerful hybrid architecture that emphasizes low-level feature learning in hybrid frameworks for multi-frame IRSTD. Our key insight is that the standard linear patch embeddings in Vision Transformers are insufficient for capturing the scale-sensitive local features critical to infrared small targets. To overcome this limitation, we design a multi-scale CNN frontend that explicitly models local features by leveraging the local spatial bias of convolution. Additionally, we introduce a U-shaped video Transformer for multi-frame spatiotemporal context modeling, effectively capturing the motion characteristics of targets. To the best of our knowledge, we are the first to introduce the concept of “low-level matters” within the CNN-Transformer hybrid architecture for multi-frame IRSTD. Despite its simplicity, with only 1.77 M parameters and 17.88 G FLOPs, LVNet achieves state-of-the-art (SOTA) performance at a speed of 208 FPS. Experiments on multiple public datasets, comparing LVNet with 18 SOTA methods, demonstrate its superior performance and efficiency. Our code and pre-trained models are accessible at https://github.com/ZhihuaShen/LVNet.