In recent years, significant advances have been gained in visual detection, and an abundance of outstanding models have been proposed. However, state-of-the-art object detection networks have some inefficiencies in detecting small targets. They commonly fail to run on portable devices or embedded systems due to their high complexity. In this workpaper, a real-time object detection model, termed as Tiny Fast You Only Look Once (TF-YOLO), is developed to implement in an embedded system. Firstly, the k-means++ algorithm is applied to cluster the dataset, which contributes to more excellent priori boxes of the targets. Secondly, inspired by the multi-scale prediction idea in the Feature Pyramid Networks (FPN) algorithm, the framework in YOLOv3 is effectively improved and optimized, by three scales to detect the earlier extracted features. In this way, the modified network is sensitive for small targets. Experimental results demonstrate that the proposed TF-YOLO method is a smaller, faster and more efficient network model increasing the performance of end-to-end training and real-time object detection for a variety of devices.

A magnetotelluric sounding has been carried out at 16 sites along a 340km long profile Yuedu‐Wuwei‐Baimagang, trending NNE across the epicentre of the 1927 M8 Gulang earthquake in Gansu Province of China. The observational data of this survey are processed by using the Robust method, yielding a series of parameters of the deep electric structure, including apparent resistivity, impedance phase, 2‐D skewness, and azimuth of principal electric axis. By means of the NLCG technique, 2‐D inversion modelling is performed to the data of two modes TE and TM. The resultant 2‐D electric structure shows that there are distinct boundaries indicative of five faults along the profile from southwest to northeast, which are the Datongshan‐Dabanshan, west Haiyuan, Huangcheng‐Shuangta, Longshoushan, and Beidashan faults. These faults become shallower and gentler from southwest to northeast, and converge into a detachment surface of low resistivity at depth, exhibiting a treeshape linkage. Along the profile, the upper crust can be divided into six electric blocks, geologically corresponding to the middle Qilian swell, north Qilian fold zone, Lenglongling swell, Wuwei basin, Chaoshui basin, and Beidashan swell, respectively. Below theWuwei basin, Chaoshui basin, and Beidashan swell, the electric structure is relatively simple in the middle and lower crust and complex in the upper crust. From low‐resistivity layers in the middle and lower crust and their depth change, it is inferred that the upper crust of the Qilian block overthrusts toward northeast while the middle and lower crust of the Alxa block underthrusts to southwest. At the contact portion between these two blocks lies the Lenglongling swell, beneath which there exists a pillar‐like body of high resistivity from the surface down to the middle and lower crust, probably implying a mountain root. The epicentre of the 1927 Gulang M8 earthquake is located at a step‐shape electric structure with notable change. South to and above the seismic source is of high resistivity, while north to and below the source is of low resistivity. In conjunction with analysis of deep electric structure and seismotectonics, it is suggested that the 1927 M8 Gulang earthquake was resulted from the overthrusting along a low‐angle detachment sheet at depth on a foreland fault under the NE‐SW directed compressive stress.

No abstract

No abstract

A 65-77 GHz low power, meander-type transmission line CMOS low-noise amplifier for E-band millimetre-wave communication using 90 nm MS/RF CMOS process is demonstrated. The low-noise amplifier consists of multi-stage cascaded common-source amplifiers with optimized noise figure, power gain, and power consumption. The LNA achieves 11.7 dB power gain, 5.5 dB noise figure, and 17.7 mW power consumption. The meander-type transmission line circuit design is adopted to reduce chip area; the total transmission line area in LNA is only 0.024 mm 2 .