针对现有行人重识别算法因细节信息关注不足导致的判别性不强问题,提出了一种基于注意力机制的多级特征级联行人重识别算法。首先,通过级联不同深度的特征实现对不同层级特征的充分利用,以补充高层级特征中的细节信息。然后,引入一对互补的注意力机制模块,以融合特征图中相似的像素及通道,弥补特征中的空间位置信息,提高特征的判别性。最后,在Market-1501、DukeMTMC-ReID、CUHK03数据集上进行了大量实验。结果表明,本算法的识别精度和平均准确率优于大多数当前的主流算法。

In practice, since the measurement environments are usually complex, the electromagnetic interference problem has been an important issue for sensor applications, and the fiber sensor can overcome it effectively. this paper proposes a new type of vibration sensor based on fiber Bragg grating (FBG). The vibration sensor with double equal strength cantilever beams structure can further improve the measurement stability. The physical model of FBG vibration sensor was established, and the structural parameters of the double equa-strength cantilever beams were designed and optimized by finite element analysis. Both simulation and experimental tests are included in this paper to illustrate the low-frequency vibration performance of the designed vibration sensor. Among, the sensitivity of the sensor was 0.024 0 pm/(m·s^-2）, the natural frequency was 185 Hz, and the linearity was obtained, which further verify the stability and reliability of the sensor vibration frequency measurement.

Fiber grating is a kind of new type of fiber optic light source device which has been rapidly changing in the refractive index of the core in recent years. Especially, it can realize the high precision of the external parameters by means of the special structure design and the encapsulation technology [1, 2]. In this paper, a fiber grating vibration sensor which is suitable for vibration monitoring in key areas is designed based on the technical background of vibration monitoring system. The sensor uses a single beam structure and pastes the fiber Bragg grating (FBG) to measure the vibration wavelength on the surface. When the vibration is simply harmonic vibration, the Bragg reflection wavelength will change periodically, and the periodic variation of the wavelength curve can be measured by the fiber grating demodulator, then the correctness of the experimental results is verified. In this paper, through the analysis of the data measured by the demodulator, the MATLAB software is used to verify the data, and the different frequency domains, the modes, and the phase frequency curves are obtained. The measurement range is 0 Hz – 100 Hz, and the natural frequency is 90.6 Hz.

In view of the problems such as the lower automation level and the insufficient precision of the traditional fiber optic gyroscope (FOG) static north-finder, this paper focuses on the in-depth analysis of the FOG dynamic north-finder principle and algorithm. The simulation model of the FOG dynamic north found algorithm with the least square method by points is established using Simulink toolbox, and then the platform rotation speed and sampling frequency, which affect FOG dynamic north found precision obviously, are simulated and calculated, and the optimization analysis is carried out as a key consideration. The simulation results show that, when the platform rotation speed is between 4.5 °/s and 8.5 °/s and the sampling frequency is at about 50 Hz in the case of using the parameters of this paper, the FOG dynamic north finding system can reach the higher precision. And the conclusions can provide the reference and validation for the engineering and practical of FOG dynamic north-finder.

In view of the existing electrical vibration monitoring traditional hydraulic pump vibration sensor, the high false alarm rate is susceptible to electromagnetic interference and is not easy to achieve long-term reliable monitoring, based on the design of a beam of the uniform strength structure of the fiber Bragg grating (FBG) vibration sensor. In this paper, based on the analysis of the vibration theory of the equal strength beam, the principle of FBG vibration tuning based on the equal intensity beam is derived. According to the practical application of the project, the structural dimensions of the equal strength beam are determined, and the optimization design of the vibrator is carried out. The finite element analysis of the sensor is carried out by ANSYS, and the first order resonant frequency is 94.739 Hz. The vibration test of the sensor is carried out by using the vibration frequency of 35 Hz and the vibration source of 50 Hz. The time domain and frequency domain analysis results of test data show that the sensor has good dynamic response characteristics, which can realize the accurate monitoring of the vibration frequency and meet the special requirements of vibration monitoring of hydraulic pump under specific environment.

In view of problems existing in the detection of the traditional hydraulic system, such as the large volume of sensor and the low measurement accuracy, a new one-piece target type flow sensor is designed and researched based on fiber Bragg grating (FBG). A compact structure is designed, which is convenient to be dismantled, processed, and installed, based on the analysis of the principle of FBG and the structure of target type flow sensor. The force of target put in fluid flow is turned into the FBG wavelength drift, with a corresponding relationship. The problem on the cross sensitivities of the temperature and strain is solved effectively by using double FBG symmetrically pasted on the both surfaces of the cantilever. The impact on the fluid state is analyzed through simulation in the software FLUENT, and the results show that the impact was smaller than that of the traditional structure. The results of experiments in the hydraulic system show that there is a good linear relationship between the change in the dual FBG central wavelength and mass loading on the target sheet has a good linear relationship, and the sensitivity is twice that of a single FBG sensitivity.