Improved SSD network for fast concealed object detection and recognition in passive terahertz security images

Cheng, Leping; Ji, Yicai; Li, Chao; Liu, Xiaojun; Fang, Guangyou

doi:10.1038/s41598-022-16208-0

Cited by 35 publications

(20 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using a convolutional backbone network, we can extract a higher-level feature map containing information of a small object at the cost of losing a lower-level feature map containing spatial information. To compensate for the trade-off, various methodologies have been proposed to combine the shallow feature with the deep feature [6][7][8][9][10][11][12][13] . These approaches can make the deeper layer contain a sufficient amount of spatial information, which is helpful in detecting small objects.…”

Section: Related Workmentioning

confidence: 99%

High-resolution processing and sigmoid fusion modules for efficient detection of small objects in an embedded system

Kim

Sung

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Recent advances in deep learning realized accurate, robust detection of various types of objects including pedestrians on the road, defect regions in the manufacturing process, human organs in medical images, and dangerous materials passing through the airport checkpoint. Specifically, small object detection implemented as an embedded system is gaining increasing attention for autonomous vehicles, drone reconnaissance, and microscopic imagery. In this paper, we present a light-weight small object detection model using two plug-in modules: (1) high-resolution processing module (HRPM ) and (2) sigmoid fusion module (SFM). The HRPM efficiently learns multi-scale features of small objects using a significantly reduced computational cost, and the SFM alleviates mis-classification errors due to spatial noise by adjusting weights on the lost small object information. Combination of HRPM and SFM significantly improved the detection accuracy with a low amount of computation. Compared with the original YOLOX-s model, the proposed model takes a two-times higher-resolution input image for higher mean average precision (mAP) using 57% model parameters and 71% computation in Gflops. The proposed model was tested using real drone reconnaissance images, and provided significant improvement in detecting small vehicles.

show abstract

Section: Related Workmentioning

confidence: 99%

High-resolution processing and sigmoid fusion modules for efficient detection of small objects in an embedded system

Kim

Sung

et al. 2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Convolutional neural networks (CNNs), a deep learning technique based on artificial neural networks, provide a powerful tool for efficient automatic object detection and recognition. Several CNN algorithms, including faster region-based CNN, 19 region-based fully convolutional networks, 20 single-shot multibox detector (SSD), 21 few-shot object detection, 22 detection transformer (DETR)-based methods, 23 , 24 and YOLO family algorithms, such as YOLOv3, 25 YOLOv5, 26 , 27 and YOLOv8, 28 have been utilized for concealed objects recognition in passive THz and sub-THz images. Most related research works focused on concealed weapon detection under clothing for personal screening.…”

Section: Introductionmentioning

confidence: 99%

Concealed objects detection and recognition in active transmission sub-terahertz imaging scanner for automatic security screening

Hocine,

Hamid,

Mohamed

2024

Opt. Eng.

View full text Add to dashboard Cite

As part of developing a sub-terahertz imaging scanner for automatic security screening applications, we investigate the detection and recognition of concealed objects in sub-terahertz images using various image processing techniques capable of addressing the low quality of such images. First, we propose a method that combines locally adaptive and multilevel thresholding with connected components labeling. This method enables the removal of the nonuniform background, facilitating the detection and localization of multiple concealed objects. Subsequently, we analyze the performance of several local feature-based matching methods for selecting the optimal detector-descriptor combination, allowing for efficient detection and recognition of concealed objects. Furthermore, we assess the applicability and performance of the supervised machine learning method based on the support vector machine classifier combined with the bag of visual words model. Finally, we explore the effectiveness of transfer learning by fine-tuning two pretrained convolutional neural networks, namely AlexNet and YOLOv3 models. The two fine-tuned models achieve recognition accuracies of 87% and 92% for AlexNet and YOLOv3, respectively. Benchmarking the study results demonstrates that deep learning methods provide an efficient solution for automatic security screening systems.

show abstract

“…Cheng et al. proposed a passive terahertz image object detection algorithm based on deep residual network, combined with multi‐scale features and mixed attention mechanism to improve detection accuracy [4]. Using image processing techniques, Sezer et al.…”

Section: Introductionmentioning

confidence: 99%

“…The proposed method completes the object detection of passive THz images with a maximum precision of 94% and a maximum recall of 90.38%. Cheng et al proposed a passive terahertz image object detection algorithm based on deep residual network, combined with multiscale features and mixed attention mechanism to improve detection accuracy [4]. Using image processing techniques, Sezer et al proposed an optimization-based deep learning model to detect solder paste defects on high-performance PCBS at an early stage [5].…”

mentioning

confidence: 99%

HE‐DETR‐DC5 method to rapidly detect objects on passive THz images in human safety inspection

Cheng

Zhang

et al. 2023

Electronics Letters

Self Cite

View full text Add to dashboard Cite

As worldwide counterterrorism operations grow, screening people for concealed objects in public places utilizing Terahertz security technology is becoming increasingly critical. In this letter, a passive THz Human security image Intelligent detection method (HE-DETR-DC5) is proposed that combines histogram equalization (HE) processing and DEtection TRansformer (DETR). The proposed method first performs histogram equalization on passive THz images to improve image contrast. In addition, the cavity convolution is added to the ResNet-50 backbone network to enhance the model's attention to the concealed object in the passive THz image and improve the accuracy. The corresponding model is called DETR-DC5. The experimental results show that the average precision mAP of the proposed method can achieve 99.9%. At the same time, compared to other image enhancement methods, the histogram equalization method used in this letter has better performance and higher detection accuracy. The detection speed of the proposed method is about 0.5 FPS, which can meet the demand for real-time detection. Therefore, the proposed model can efficiently and automatically detect hidden objects in passive THz human security images.

show abstract

Improved SSD network for fast concealed object detection and recognition in passive terahertz security images

Cited by 35 publications

References 33 publications

High-resolution processing and sigmoid fusion modules for efficient detection of small objects in an embedded system

High-resolution processing and sigmoid fusion modules for efficient detection of small objects in an embedded system

Concealed objects detection and recognition in active transmission sub-terahertz imaging scanner for automatic security screening

HE‐DETR‐DC5 method to rapidly detect objects on passive THz images in human safety inspection

Contact Info

Product

Resources

About