Dim and Small Target Detection Based on Local Energy Aggregation Degree of Sequence Images

Fan, Xiangsuo; Xu, Zhihai

doi:10.1155/2019/9282141

Cited by 4 publications

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The simulation results indicated that the identification precision through the fusion of MMW radar and the camera significantly outperformed that of single sensor. An anisotropic gradient background modelling method combined with spatial and temporal information was presented [9] to improve the dim and small targets identification in the dynamic scenes. The multidirectional gradient maximum of neighboring blocks was used to segment the difference maps.…”

Section: Past Workmentioning

confidence: 99%

Target Identification Based on Intensity, Texture, and Sparse Dictionary

Azizi

Mohanna

Ahsani

2023

Preprint

View full text Add to dashboard Cite

It is very difficult to inspect by human at large borders and in difficult border crossings. Despite much research into the target identification, there are still challenges that make target identification difficult in the borders. Because in the borders, there is more similarity between the target and the background, and usually equipment in the borders uses maximum camouflage. This paper attempts to create an intelligent target identification software for the remote control system to identify targets based on the intensity, texture, and sparse dictionary. The input image is divided into the super-pixels by using the simple linear iterative clustering algorithm. To obtain sufficient information, the standard intensity features and Gabor texture are extracted from each super-pixel in the frequency domain. To identify the targets, several background sparse dictionaries are created. The super-pixels and the fuzzy C-means clustering are used to construct the initial dictionary. By assigning the super-pixels with the sparser representation in a dictionary, a new class is created for each dictionary. Then, these classes are used to update dictionaries. The targets are identified based on the combination of coding errors and representation coefficients. The simulation results are obtained on a database prepared by the authors. The simulation results are evaluated using Dice, specificity, sensitivity and accuracy criteria. According to the criteria, the proposed method has more successful performance than the traditional sparse representation classification method. The final performance of the proposed method is 96.8%.

show abstract

Section: Past Workmentioning

confidence: 99%