IRSDet: Infrared Small-Object Detection Network Based on Sparse-Skip Connection and Guide Maps

Abstract

Detecting small objects in infrared images remains a challenge because most of them lack shape and texture. In this study, we proposed an infrared small-object detection method to improve the capacity for detecting thermal objects in complex scenarios. First, a sparse-skip connection block is proposed to enhance the response of small infrared objects and suppress the background response. This block is used to construct the detection model backbone. Second, a region attention module is designed to emphasize the features of infrared small objects and suppress background regions. Finally, a batch-averaged biased classification loss function is designed to improve the accuracy of the detection model. The experimental results show that the proposed small-object detection framework significantly increases precision, recall, and F1-score, showing that, compared with the current advanced detection models for small-object detection, the proposed detection framework has better performance in infrared small-object detection under complex backgrounds. The insights gained from this study may provide new ideas for infrared small object detection and tracking.

IRSDT: A Framework for Infrared Small Target Tracking with Enhanced Detection

“…Additionally, they designed a dual-branch parallel structure at the neck of the network, with one branch serving as the pooling downsampling branch to further obtain the target's overall semantic information and the other as a branch to enable complete transmission of the small target information to the deep layers. Xi et al [26] proposed a guided map module similar to a region spatial attention mechanism, which uses the characteristics of the infrared target's high grayscale responses to locate and enhance the grayscale features of the potential areas of the infrared small targets.…”

“…Researchers are currently providing many optimization proposals for enhancing the detection ability of the CNN-based infrared small target detection model. Several studies have analyzed the characteristics and motion states of infrared small targets and optimized the infrared small target detection algorithm in areas such as the spatial-temporal information [15][16][17], multiscale information [18][19][20][21], feature enhancement [22][23][24][25][26], and training and detection strategies [27,28], achieving relatively good detection results. However, most methods do not consider computing costs and parameter reduction in small models, limiting the likelihood of the CNN-based infrared small target detection model being used in edge computing hardware platforms.…”

Small object detection based on hierarchical attention mechanism and multi‐scale separable detection

“…There remains a gap between the theoretical receptive field and the actual receptive field, posing difficulties in handling target deformations [18,19]. Some algorithms [20][21][22] employ an attention mechanism to tackle occlusion, suppressing extraneous noise interference and significantly improving detection accuracy. However, these algorithms encounter difficulties in effectively capturing global image feature dependencies while simultaneously emphasizing local features.…”

Infrared image-based remote target detection for maritime rescue utilizing a deep learning network and data augmentation

“…Compared to conventional RGB images, IR imaging has advantages in object detection tasks owing to its illumination and environmental robustness. However, IR images inherently suffer from low contrast and ambiguous textile feature, which contribute to difficulties in detecting and localizing remote targets 2 . Additionally, IR image dataset containing high-resolution remote targets is scarce in general which makes the training-based automation of object detection very difficult.…”