Infrared Small Target Detection Method Based on Curvature near the Ground

Guo-qiang, 朱国强 Zhu; Xiang-yong, 孟祥勇 Meng; Weixian, 钱惟贤 Qian

doi:10.3788/gzxb20184710.1010001

Cited by 6 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to the above analysis, the curvature feature point extraction algorithm for point target [4] is designed, and the algorithm flow chart is shown in the following Figure 2.…”

Section: Algorithm Designmentioning

confidence: 99%

Research on multi-scale moving target detection method under mobile platform

Lin

Gao²,

Qian

2023

AOPC 2022: Optical Sensing, Imaging, and Display Technology

View full text Add to dashboard Cite

This paper firstly obtains a set of images with different resolutions by down-sampling the original image, so that there is always a point target in the set of images for targets of different scales. Then, according to the characteristic that the point target exhibits a Gaussian-like distribution, a curvature-based filtering algorithm is used for the group of images to suppress the background and extract the feature points. For the feature points extracted stably from multiple frames, they are correlated to form tracks. Aiming at the problem that the motion law of background points is difficult to be counted due to the insufficient number of background points in low resolution images. This paper counts and analyzes the relationship between the track velocity of background points in images with different resolutions. It is found that the average velocity fluctuation of the background points in the images with different resolutions is basically the same, and has a multiple relationship. According to this relationship, this paper adopts the method of velocity mapping to multiply the velocity of the track in the low-resolution image by the multiple relationship, and perform unified clustering with the track in the highresolution image. This ensures that the background point track has a large number of samples and improves the stability of the clustering under the premise that the background point track has the same motion law. Through the unified clustering of all tracks, the target tracks at each resolution can be finally screened out.

show abstract

“…According to the above analysis, the curvature feature point extraction algorithm for point target [4] is designed, and the algorithm flow chart is shown in the following Figure 2.…”

Section: Algorithm Designmentioning

confidence: 99%