Multi-Spectral RGB-NIR Image Classification Using Double-Channel CNN

Jiang, Jionghui; Feng, Xiaoyi; Fen, Liu; Xu, Yingying; Huang, Hui

doi:10.1109/access.2019.2896128

Cited by 64 publications

(28 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In future work, we will focus on (i) considering other CNN models [29] and image features, (ii) considering a more accurate method to obtain the query class.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A New Re-Ranking Method Based on Convolutional Neural Network and Two Image-to-Class Distances for Remote Sensing Image Retrieval

Dong

Luo

et al. 2019

IEEE Access

View full text Add to dashboard Cite

With the rapid growth of remote sensing image data, it has become necessary to effectively and efficiently retrieve images from a big image database for managing and exploiting such data. This paper presents a novel method for content-based remote sensing image retrieval (CBRSIR) that re-ranks the initial retrieval result using two image-to-class distances, which are the similarity between an image and an image class. One is the image-to-training-class distance between an image and each image class of the training dataset, which uses the information included in the train samples with known-label. It is used to calculate the weight of the image classes and estimating the class of an image. The other is the image-to-queryclass distance between an image and the query class that is obtained using the k-nearest neighbor (kNN) method. The proposed method first obtains the initial retrieval results via sorting the distances of the CNN feature between the query image and each retrieved image in ascending order. Then, the image-to-query-class distance is calculated according to the initial result, and the weight of each class is calculated according to the class probability of images obtained by CNN and the image-to-training-class distances. Finally, the initial result is re-ranked to get the final retrieval result via the weighted image-to-class distances. The performance of the proposed method is tested on UCMD and PatternNet databases. Experimental results show the effectiveness of the proposed re-ranking scheme in image retrieval in comparison to other state-of-the-art techniques.INDEX TERMS Remote sensing image retrieval (RSIR), re-ranking, convolutional neural network (CNN), image-to-class distance. I. INTRODUCTIONWith the rapid development of technology of remote sensing (RS) sensors, a large number of high-resolution remote sensing (HRRS) images are acquired from satellites and aircraft [1]. Content-based remote sensing image retrieval (CBRSIR) is one of the most fundament techniquesThe associate editor coordinating the review of this manuscript and approving it for publication was Maurizio Tucci.

show abstract

“…In future work, we will focus on (i) considering other CNN models [29] and image features, (ii) considering a more accurate method to obtain the query class.…”

Section: Discussionmentioning

confidence: 99%

“…CNN, a hierarchical neural network, can extract higherlevel features and classify objects at the same time, which has already been used in various fields such as image classification [28], [29], and image scene analysis [30]. We select the VGG [31] and the ResNet Model [32] to extract features for RSIR in the proposed method.…”

Section: A Feature Extractionmentioning

confidence: 99%

A New Re-Ranking Method Based on Convolutional Neural Network and Two Image-to-Class Distances for Remote Sensing Image Retrieval

Dong

Luo

et al. 2019

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Among these methods, detection system based on learning is the best because it achieves a detection rate to more than 90%. Using a deep learning method such as convolutional neural network (CNN) even increases the detection rate to more than 98% [6]. However, the existing algorithms are still far from a real time classified system [4].…”

Section: Previous Workmentioning

confidence: 99%

Traffic sign detection optimization using color and shape segmentation as pre-processing system

Handoko¹,

Pratama²,

Yohanes³

2021

TELKOMNIKA

View full text Add to dashboard Cite

One of performance indicators in an autonomous vehicle (AV) is its ability to accommodate rapid environment changing; and performance of traffic sign detection (TSD) system is one of them. A low frame rate of TSD impacts to late decision making and may cause to a fatal accident. Meanwhile, adding any GPU to TSD will significantly increases its cost and make it unaffordable. This paper proposed a pre-processing system for TSD which implement a color and a shape segmentation to increase the system speed. These segmentation systems filter input frames such that the number of frames sent to artificial intelligence (AI) system is reduced. As a result, workload of AI system is decreased and its frame rate increases. HSV threshold is used in color segmentation to filter frames with no desired color. This algorithm ignores the saturation when performing color detection. Further, an edge detection feature is employed in shape segmentation to count the total contours of an object. Using German traffic sign recognition benchmark dataset as model, the pre-processing system filters 97% of frames with no traffic sign objects and has an accuracy of 88%. TSD system proposed allows a frame rate improvement up to 32 frame per second (FPS) when You Only Look Once (YOLO) algorithm is used. This is an open access article under the CC BY-SA license.

show abstract

“…Meanwhile, the feature combination strategy was also applied in this study to improve the performance of the model because of its sound effects in the model establishment processes. 21…”

Section: Introductionmentioning

confidence: 99%

Fractional order modeling and recognition of nitrogen content level of rubber tree foliage

Tang

et al. 2020

Journal of Near Infrared Spectroscopy

View full text Add to dashboard Cite

The Nondestructive estimation method of nitrogen content level of rubber tree foliage was investigated utilizing nearinfrared (NIR) spectroscopy and Grünwald-Letnikov fractional calculus. Four models, including partial least squares discriminant analysis (PLS-DA), support vector machine (SVM), extreme learning machine (ELM) and convolutional neural networks (CNN) are applied to construct the nitrogen estimation model. The results show that models established by 0.6-order or 1.6-order spectra achieved better performance than models with integer-order spectra. Afterward, the successive projections algorithm (SPA) is applied to reduce the number of variables, which is critical for developing portable nitrogen-level detector devices for rubber trees. The PLS-DA method achieved the best performance with an optimal recognition rate (97.73%) using the 1.6-order spectra. The results suggest that nitrogen content of rubber trees could be reliably estimated by fractional calculus processed NIR spectra. The method proposed here has a wide range of applicability and can provide more useful information for NIR spectral analysis in agriculture as well as other fields.

show abstract

Multi-Spectral RGB-NIR Image Classification Using Double-Channel CNN

Cited by 64 publications

References 26 publications

A New Re-Ranking Method Based on Convolutional Neural Network and Two Image-to-Class Distances for Remote Sensing Image Retrieval

A New Re-Ranking Method Based on Convolutional Neural Network and Two Image-to-Class Distances for Remote Sensing Image Retrieval

Traffic sign detection optimization using color and shape segmentation as pre-processing system

Fractional order modeling and recognition of nitrogen content level of rubber tree foliage

Contact Info

Product

Resources

About