Smoking behavior in public places and fire bans seriously threatens the safety of people's lives and property. In order to ensure public safety, this paper proposes a smoking behavior detection method based on YOLOv5 algorithm and image processing. Aiming at the small target of cigarettes, this paper uses the K-means algorithm and the method of adding a small target detection layer to improve the YOLOv5 algorithm, and realizes the improvement of the detection accuracy. On the self-made data set, the false detection rate is 0%, and the AP is 92.3%, which is 6.7% higher than that of the YOLOv5s algorithm.

A novel self-calibration comparator for a 12-bit 2.5 MSPS successive approximation register analog-to-digital converter (SAR ADC) applied in a touch microcontroller unit (MCU) with small area, high precision, fast response speed, and low-voltage detection is proposed in this paper. A combination of input/output offset storage (IOS/OOS) and an offset trimming circuit was employed to reduce the offset of the cascade preamplifier and the operational transconductance amplifier (OTA), a novel offset trimming circuit with a 5-bit digital controller was designed to further reduce the residual offset voltage, and an improved self-calibration technology was also implemented to compensate the conversion error in SAR ADC system to a minimum. Simulation and measured results show that the input-referred offset calibrating range is ±9.15 mV at 0.61 mV/step, the low-voltage detection of SAR ADC is realized by compensating the conversion error to a minimum, and the effective number of bits (ENOB) and figure of merit (FoM) at 5 V supply and 2.5 M/s rate in the 12-bit SAR ADC with a 95 nm CMOS are 11.33 bits and 726.6 fJ/conversion-step, respectively. The proposed self-calibration comparator applied in the SAR ADC system can automatically eliminate the offset voltage caused by nonidealities and meet the requirements of the touch MCU.

Surface defects, which often occur during the production of aluminum profiles, can directly affect the quality of aluminum profiles, and should be monitored in real time. This paper proposes an effective, lightweight detection method for aluminum profiles to realize real-time surface defect detection with ensured detection accuracy. Based on the YOLOv5s framework, a lightweight network model is designed by adding the attention mechanism and depth-separable convolution for the detection of aluminum. The lightweight network model improves the limitations of the YOLOv5s framework regarding to its detection accuracy and detection speed. The backbone network GCANet is built based on the Ghost module, in which the Attention mechanism module is embedded in the AC3Ghost module. A compression of the backbone network is achieved, and more channel information is focused on. The model size is further reduced by compressing the Neck network using a deep separable convolution. The experimental results show that, compared to YOLOv5s, the proposed method improves the mAP by 1.76%, reduces the model size by 52.08%, and increases the detection speed by a factor of two. Furthermore, the detection speed can reach 17.4 FPS on Nvidia Jeston Nano’s edge test, which achieves real-time detection. It also provides the possibility of embedding devices for real-time industrial inspection.

An adaptive high-efficiency light-emitting Diode (LED) backlight driver scheme has been proposed to address the issue of additional power loss caused by LED forward voltage variation. In this scheme, the peak current and the duty cycle of each LED channel are adjusted separately through an adaptive control algorithm to minimize the voltage drop on the linear current regulator (LCR) of each LED channel to reduce the excessive power loss in each LED channel and enhance the total power efficiency. A linear current regulator, suitable for adaptive control, is designed on a 0.18 μm 5V complementary metal-oxide-semiconductor (CMOS) process. Simulation results demonstrate that the linear current regulator can achieve a linearly adjustable channel current ranging from 0 to 48 mA with a current resolution of 0.2 mA. Across different process corners and temperatures, the maximum error for the full current range is less than 0.1%. The core area of chip layout is about 0.1 mm². The complete driver prototype comprises the LCR chips, external power MOS transistors, digital module, and LED chains. The test results show that the power loss of the linear current regulator has been significantly reduced, and the power efficiency of each LED channel has been measured at around 98.1%.

4D mmWave radar-vision fusion is a low-cost and promising solution for Unmanned Surface Vehicles navigation. However, existing methods mainly focus on single tasks and suffer from insufficient radar feature extraction and inefficient multi-modal fusion. This paper proposes RVFNet, a 4D mmWave radar-visual fusion multi-task perception network. It designs the Radar Feature Enhancement Module (RFEM) and alternately stacks RFEM with conventional convolution blocks to construct the Radar Feature Extraction Network (RFENet), which solves the problem of insufficient extraction of sparse and irregular radar point cloud features and suppresses water-surface clutter; it constructs the Multi-modal Feature Adaptive Fusion Module (MFAFM) with content-guided attention. MFAFM generates dynamic modal weights and adaptively balances visual and radar information to alleviate missed detection and false detection in harsh scenarios. Compared with mainstream methods, experimental results on the WaterScenes dataset show that RVFNet requires only 3.3 GFLOPs while achieving an mAP${}_{50}$ of 79.0%, an mAP${}_{50--95}$ of 46.7%, an mIoU of 84.0% for object segmentation, and an mIoU of 98.8% for drivable-area segmentation, demonstrating superior overall perception performance. Cross-scenario experiments on the FloW-RI water-surface dataset and the nuScenes v1.0-trainval autonomous-driving dataset further verify the strong generalization capability of RVFNet, demonstrating its potential for deployment on embedded devices.

Due to the lack of precise models for the multi-layer helical inductors, the actual performance of the vertical T-coil broadband ESD protection circuit could deviate from what estimated at the design stage. In this paper, we present an improved broadband ESD protection circuit with tunability and investigate the effectiveness of the capacitive tuning structure on improving and optimizing the frequency characteristics of the circuit.