Design and Training of Deep CNN-Based Fast Detector in Infrared SUAV Surveillance System

Zhang, Yunlong; Zhang, Yan; Shi, Zhongke; Jing-hua, Zhang; Wei, Ming

doi:10.1109/access.2019.2941509

Cited by 23 publications

(12 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When used in the infrared system, diversities between characteristics of images may cause more problems, making recognition more challenging. In [44], a DCNN-based detector was designed in the infrared small unmanned aerial vehicle (SUAV) surveillance system by the laterally connected multi-scale feature fusion approach and densely paved predefined boxes. In [45], SVM and DCNN classification for infrared target recognition were compared.…”

Section: Dcnn-based Target Recognitionmentioning

confidence: 99%

Target Recognition in Infrared Circumferential Scanning System via Deep Convolutional Neural Networks

Chen

2020

Sensors

View full text Add to dashboard Cite

With an infrared circumferential scanning system (IRCSS), we can realize long-time surveillance over a large field of view. Recognizing targets in the field of view automatically is a crucial component of improving environmental awareness under the trend of informatization, especially in the defense system. Target recognition consists of two subtasks: detection and identification, corresponding to the position and category of the target, respectively. In this study, we propose a deep convolutional neural network (DCNN)-based method to realize the end-to-end target recognition in the IRCSS. Existing DCNN-based methods require a large annotated dataset for training, while public infrared datasets are mostly used for target tracking. Therefore, we build an infrared target recognition dataset to both overcome the shortage of data and enhance the adaptability of the algorithm in various scenes. We then use data augmentation and exploit the optimal cross-domain transfer learning strategy for network training. In this process, we design the smoother L1 as the loss function in bounding box regression for better localization performance. In the experiments, the proposed method achieved 82.7 mAP, accomplishing the end-to-end infrared target recognition with high effectiveness on accuracy.

show abstract

Section: Dcnn-based Target Recognitionmentioning

confidence: 99%

Target Recognition in Infrared Circumferential Scanning System via Deep Convolutional Neural Networks

Chen

2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Many applications of machine learning and, most recently, computer vision have been disrupted by the use of CNNs [ 11 , 12 , 13 , 14 , 15 ]. Combining a minimal need for human design and the efficient training of large and complex models has allowed them to achieve state-of-the-art performance on several benchmarks.…”

Section: Introductionmentioning

confidence: 99%

“…Due to this, techniques such as transfer learning [ 26 ] have become popular, utilizing the application of architectures that have been trained previously to solve new classification problems. These models can effectively serve as a generic model of the visual world, with different approximations [ 11 , 12 , 13 , 15 ].…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Architecture Reduction for fMRI Data

Alvarez-Gonzalez

Méndez-Vázquez

2022

Brain Sciences

View full text Add to dashboard Cite

In recent years, deep learning models have demonstrated an inherently better ability to tackle non-linear classification tasks, due to advances in deep learning architectures. However, much remains to be achieved, especially in designing deep convolutional neural network (CNN) configurations. The number of hyper-parameters that need to be optimized to achieve accuracy in classification problems increases with every layer used, and the selection of kernels in each CNN layer has an impact on the overall CNN performance in the training stage, as well as in the classification process. When a popular classifier fails to perform acceptably in practical applications, it may be due to deficiencies in the algorithm and data processing. Thus, understanding the feature extraction process provides insights to help optimize pre-trained architectures, better generalize the models, and obtain the context of each layer’s features. In this work, we aim to improve feature extraction through the use of a texture amortization map (TAM). An algorithm was developed to obtain characteristics from the filters amortizing the filter’s effect depending on the texture of the neighboring pixels. From the initial algorithm, a novel geometric classification score (GCS) was developed, in order to obtain a measure that indicates the effect of one class on another in a classification problem, in terms of the complexity of the learnability in every layer of the deep learning architecture. For this, we assume that all the data transformations in the inner layers still belong to a Euclidean space. In this scenario, we can evaluate which layers provide the best transformations in a CNN, allowing us to reduce the weights of the deep learning architecture using the geometric hypothesis.

show abstract

“…However, due to the small image size, only gray information and unclear features, the existing target detection algorithm based on deep learning in the field of machine vision is not suitable for infrared small target detection [4] . Aiming at the high false alarm rate of small targets, Xu et al [5] redesigned the SSD network structure by adjusting the high and low resolution layers; Article [6] uses regression type deep convolution neural network to suppress background components, and extracts candidate target regions through threshold segmentation; Shi et al [7] regard the small target in the infrared image as "noise", transform the small target detection task into denoising problem, and use the perceptual loss to solve the loss of background texture features in the coding process, which can realize the unsupervised feature extraction of invariant features and adapt to the infrared small target detection task in different fields. Aiming at the problem of target detection under the situation of target occlusion, deformation and incomplete exposure of target parts, Zhang et al [8] defined the loss function with constraints on the basis of the analysis of target structure and infrared characteristics of air aircraft, constructed the deep convolution network model for air target key part recognition, and proposed an air infrared target key part detection algorithm based on key point detection convolution network It shows superior performance in robustness.…”

Section: Introductionmentioning

confidence: 99%

Infrared target detection method using adaptive feature fusion

Tongtong

Peng

et al. 2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

As the key technology of low altitude airspace security and air attack, vision-based air target detection has better concealment and lower cost compared with radar, radio and other detection methods. Based on yolov3, this paper proposes an air infrared target detection algorithm using adaptive feature fusion, which can effectively improve the accuracy of air infrared multi-scale target detection. Firstly, the construction of air infrared target detection data set is completed, and the target characteristics in the data set are statistically analysed. Secondly, the output layer of backbone is adjusted, and the four times down sampling feature map is used to replace the eight times down sampling feature map to enhance the detection ability of small targets. Thirdly, aiming at the problem of weakening feature expression caused by cross-scale feature fusion in FPN, a pixel level feature adaptive fusion module is designed. Finally, two improved SPP modules are added to neck to enrich the expression ability of feature map through the fusion of local features and global features. Experimental results show that compared with yolov3, our algorithm improves the map from 86.53% to 90.26% while keeping the detection speed basically unchanged.

show abstract

Design and Training of Deep CNN-Based Fast Detector in Infrared SUAV Surveillance System

Cited by 23 publications

References 32 publications

Target Recognition in Infrared Circumferential Scanning System via Deep Convolutional Neural Networks

Target Recognition in Infrared Circumferential Scanning System via Deep Convolutional Neural Networks

Deep Learning Architecture Reduction for fMRI Data

Infrared target detection method using adaptive feature fusion

Contact Info

Product

Resources

About