ABSTRACT
Pig tracking provides strong support for refined management in pig farms. However, long and continuous multi-pig tracking is still extremely challenging due to occlusion, distortion, and motion blurring in real farming scenarios. This study proposes a long-term video tracking method for group-housed pigs based on improved StrongSORT, which can significantly improve the performance of pig tracking in production scenarios. In addition, this research constructs a 24 h pig tracking video dataset, providing a basis for exploring the effectiveness of long-term tracking algorithms. For object detection, a lightweight pig detection network, YOLO v7-tiny_Pig, improved based on YOLO v7-tiny, is proposed to reduce model parameters and improve detection speed. To address the target association problem, the trajectory management method of StrongSORT is optimized according to the characteristics of the pig tracking task to reduce the tracking identity (ID) switching and improve the stability of the algorithm. The experimental results show that YOLO v7-tiny_Pig ensures detection applicability while reducing parameters by 36.7% compared to YOLO v7-tiny and achieving an average video detection speed of 435 frames per second. In terms of pig tracking, Higher-Order Tracking Accuracy (HOTA), Multi-Object Tracking Accuracy (MOTP), and Identification F1 (IDF1) scores reach 83.16%, 97.6%, and 91.42%, respectively. Compared with the original StrongSORT algorithm, HOTA and IDF1 are improved by 6.19% and 10.89%, respectively, and Identity Switch (IDSW) is reduced by 69%. Our algorithm can achieve the continuous tracking of pigs in real scenarios for up to 24 h. This method provides technical support for non-contact pig automatic monitoring.
ABSTRACT
The authors have designed a novel type of periodic rectangular pit nanostructure substrate based on the surface plasmon principle. Finite element method was employed to simulate the optical near-field distribution. Strongly enhanced field whose electric intensity Emax/E0 can be as high as 20 at resonance frequency appears around pithead of the periodic structure. As the period of structure, pit length l, width w and environment change, the authors observe the regular shifting of plasmon resonant wavelength which can cover the range from 500 to 1000 nm. The red shifts of SPR resonance peaks are increased with the increment of period Px when incident light is polarized along x axis. An abrupt decrease in localized electric field in the pit is observed as incident wavelength approaches Px. This is due to the satisfaction of wave vector matching condition and the excitation of propagating SPP. SPR resonance peaks also red shifts with the increment on pit length l and environment dielectric refractive index, presenting a linear dependence with pit length l. While the resonance peaks are blue shifted with the increment of pit width w. The results presented in this paper will provide a way to tune the plasmon resonant wavelength. Inspired by Jain's report, SPR resonance peaks' shifting with the changing of structure parameters can be explained by viewing the rectangular pit nanostructure as combination of two pairs of dipole-dipole coupling models along x and y axis respectively.
ABSTRACT
In order to solve the problem that CCDs cannot measure the full spectral range in a single measurement, we propose a new wavelength-fitting algorithm that combines the polynomial algorithm applied to the fixed grating with grating equation without CCD or spectrum assembling. Both the grating rotating angle and pixel coordinate of the CCD are written in our wavelength-fitting function. With the calibration of the 576.96 and 579.07 nm mercury spectral line, we can determine that wavelength error of 576.96 nm is between 0.002 and 0.1 nm and wavelength error of 579.07 nm is between 0.006 and 0.06 nm. The calculation results show that the new algorithm can gain more precise wavelength accuracy without a complex assembling operation.
