Jun Ling, Hongxing Wu, A small target motion detection algorithm in complex dynamic environment
Full Text: PDF
DOI: 10.23952/jano.6.2024.3.07
Volume 6, Issue 3, 1 December 2024, Pages 411-427
Abstract. Small object motion detection in complex dynamic environments has long been a challenge in computer vision due to the limited visual features of small objects and the presence of numerous fake features in the complex background. Biological studies revealed a specialized class of neurons in the insect brain, known as small target motion detectors (STMDs), which possess the remarkable ability to flawlessly detect small object motion within the visual field. Inspired by this remarkable biological discovery, researchers proposed various small object motion detection visual networks that demonstrate promising performance in detecting small object motion. However, these visual networks lack the capability to effectively filter out background fake features, which leads to a significant number of fake features in their detection results. To address this challenge, in this paper, we propose a novel visual neural network inspired by the insect visual system and the differential responses of STMD neurons to targets and background fake feature, capable of detecting small objects and eliminating background fake features. Our visual network primarily consists of two stages: a motion information processing stage and a response discrimination stage. The motion information processing stage detects object motion by extracting object motion information, while the response discrimination stage discriminates between small objects and background fake features by utilizing the response information from the motion information processing stage. Experimental results demonstrate that our visual network successfully filters out background false positives and performs significantly better in detecting small targets in complex dynamic backgrounds.
How to Cite this Article:
J. Ling, H. Wu, A small target motion detection algorithm in complex dynamic environment, J. Appl. Numer. Optim. 6 (2024), 411-427.