• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.6
• /
• pp.1099-1110
• /
• 2023

The recognition system of autonomous driving and robot navigation performs vision work such as object recognition, tracking, and lane detection after multi-sensor fusion to improve performance. Currently, research on a deep learning model based on the fusion of a camera and a lidar sensor is being actively conducted. However, deep learning models are vulnerable to adversarial attacks through modulation of input data. Attacks on the existing multi-sensor-based autonomous driving recognition system are focused on inducing obstacle detection by lowering the confidence score of the object recognition model.However, there is a limitation that an attack is possible only in the target model. In the case of attacks on the sensor fusion stage, errors in vision work after fusion can be cascaded, and this risk needs to be considered. In addition, an attack on LIDAR's point cloud data, which is difficult to judge visually, makes it difficult to determine whether it is an attack. In this study, image scaling-based camera-lidar We propose an attack method that reduces the accuracy of LCCNet, a fusion model (camera-LiDAR calibration model). The proposed method is to perform a scaling attack on the point of the input lidar. As a result of conducting an attack performance experiment by size with a scaling algorithm, an average of more than 77% of fusion errors were caused.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.6
• /
• pp.1833-1848
• /
• 2022

Named entity recognition (NER) is an important basic task in the field of Natural Language Processing (NLP). Recently deep learning approaches by extracting word segmentation or character features have been proved to be effective for Chinese Named Entity Recognition (CNER). However, since this method of extracting features only focuses on extracting some of the features, it lacks textual information mining from multiple perspectives and dimensions, resulting in the model not being able to fully capture semantic features. To tackle this problem, we propose a novel Multi-view Semantic Feature Fusion Model (MSFM). The proposed model mainly consists of two core components, that is, Multi-view Semantic Feature Fusion Embedding Module (MFEM) and Multi-head Self-Attention Mechanism Module (MSAM). Specifically, the MFEM extracts character features, word boundary features, radical features, and pinyin features of Chinese characters. The acquired font shape, font sound, and font meaning features are fused to enhance the semantic information of Chinese characters with different granularities. Moreover, the MSAM is used to capture the dependencies between characters in a multi-dimensional subspace to better understand the semantic features of the context. Extensive experimental results on four benchmark datasets show that our method improves the overall performance of the CNER model.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.12
• /
• pp.1-7
• /
• 2019

With the in-depth development of machine learning, the deep learning method has made great progress, especially with the Convolution Neural Network(CNN). Compared with traditional text sentiment classification methods, deep learning based CNNs have made great progress in text classification and processing of complex multi-label and multi-classification experiments. However, there are also problems with the neural network for text sentiment classification. In this paper, we propose a fusion model based on Long-Short Term Memory networks(LSTM) and CNN deep learning methods, and applied to multi-category news datasets, and achieved good results. Experiments show that the fusion model based on deep learning has greatly improved the precision and accuracy of text sentiment classification. This method will become an important way to optimize the model and improve the performance of the model.

• Journal of Korea Multimedia Society
• /
• v.25 no.6
• /
• pp.819-832
• /
• 2022

Cross-domain recommender system transfers knowledge across different domains to improve the recommendation performance in a target domain that has a relatively sparse model. However, they suffer from the "negative transfer" in which transferred knowledge operates as noise. This paper proposes a novel Multiple Fusion-based Deep Cross-Domain Recommendation named MFDCR. We exploit Doc2Vec, one of the famous word embedding techniques, to fuse data user-wise and transfer knowledge across multi-domains. It alleviates the "negative transfer" problem. Additionally, we introduce a simple multi-layer perception to learn the user-item interactions and predict the possibility of preferring items by users. Extensive experiments with three domain datasets from one of the most famous services Amazon demonstrate that MFDCR outperforms recent single and cross-domain recommendation algorithms. Furthermore, experimental results show that MFDCR can address the problem of "negative transfer" and improve recommendation performance for multiple domains simultaneously. In addition, we show that our approach is efficient in extending toward more domains.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.6
• /
• pp.635-644
• /
• 2020

Object recognition, detection and instance segmentation based on DL (Deep Learning) have being used in various practices, and mainly optical images are used as training data for DL models. The major objective of this paper is object segmentation and building detection by utilizing multimodal datasets as well as optical images for training Detectron2 model that is one of the improved R-CNN (Region-based Convolutional Neural Network). For the implementation, infrared aerial images, LiDAR data, and edges from the images, and Haralick features, that are representing statistical texture information, from LiDAR (Light Detection And Ranging) data were generated. The performance of the DL models depends on not only on the amount and characteristics of the training data, but also on the fusion method especially for the multimodal data. The results of segmenting objects and detecting buildings by applying hybrid fusion - which is a mixed method of early fusion and late fusion - results in a 32.65% improvement in building detection rate compared to training by optical image only. The experiments demonstrated complementary effect of the training multimodal data having unique characteristics and fusion strategy.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.11
• /
• pp.4557-4572
• /
• 2020

The Alaska competition provides an opportunity to study the practical problems of real-world steganalysis. Participants are required to solve steganalysis involving various embedding schemes, inconsistency JPEG Quality Factor and various processing pipelines. In this paper, we propose a method to ensemble multiple deep learning steganalyzers. We select SRNet and RESDET as our base models. Then we design a three-layers model ensemble network to fuse these base models and output the final prediction. By separating the three colors channels for base model training and feature replacement strategy instead of simply merging features, the performance of the model ensemble is greatly improved. The proposed method won second place in the Alaska 1 competition in the end.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.39 no.5
• /
• pp.289-295
• /
• 2021

Recently, owing to the development of autonomous vehicles, research on precisely determining the position of a moving object has been actively conducted. Previous research mainly used the fusion of GNSS/IMU (Global Positioning System / Inertial Navigation System) and sensors attached to the vehicle through a Kalman filter. However, in recent years, new technologies have been used to determine the location of a moving object owing to the improvement in computing power and the advent of deep learning. Various techniques using RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), and NARX (Nonlinear Auto-Regressive eXogenous model) exist for such learning-based positioning methods. The purpose of this study is to compare the precision of existing filter-based sensor fusion technology and the NARX-based method in case of GNSS signal blockages using simulation data. When the filter-based sensor integration technology was used, an average horizontal position error of 112.8 m occurred during 60 seconds of GNSS signal outages. The same experiment was performed 100 times using the NARX. Among them, an improvement in precision was confirmed in approximately 20% of the experimental results. The horizontal position accuracy was 22.65 m, which was confirmed to be better than that of the filter-based fusion technique.

• Journal of Information Processing Systems
• /
• v.19 no.5
• /
• pp.673-687
• /
• 2023

Convolutional neural network-based deep learning technology is the most commonly used in image identification, but it requires large-scale data for training. Therefore, application in specific fields in which data acquisition is limited, such as in the military, may be challenging. In particular, the identification of ground weapon systems is a very important mission, and high identification accuracy is required. Accordingly, various studies have been conducted to achieve high performance using small-scale data. Among them, the ensemble method, which achieves excellent performance through the prediction average of the pre-trained models, is the most representative method; however, it requires considerable time and effort to find the optimal combination of ensemble models. In addition, there is a performance limitation in the prediction results obtained by using an ensemble method. Furthermore, it is difficult to obtain the ensemble effect using models with imbalanced classification accuracies. In this paper, we propose a transfer learning-based feature fusion technique for heterogeneous models that extracts and fuses features of pre-trained heterogeneous models and finally, fine-tunes hyperparameters of the fully connected layer to improve the classification accuracy. The experimental results of this study indicate that it is possible to overcome the limitations of the existing ensemble methods by improving the classification accuracy through feature fusion between heterogeneous models based on transfer learning.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.416-418
• /
• 2022

Glioma is a type of brain tumor that occurs in glial cells and is classified into two types: high hrade hlioma with a poor prognosis and low grade glioma. Magnetic resonance imaging (MRI) as a non-invasive method is widely used in glioma diagnosis research. Studies to obtain complementary information by combining multiple modalities to overcome the incomplete information limitation of single modality are being conducted. In this study, we developed a 3D CNN-based model that applied input-level fusion to MRI of four modalities (T1, T1Gd, T2, T2-FLAIR). The trained model showed classification performance of 0.8926 accuracy, 0.9688 sensitivity, 0.6400 specificity, and 0.9467 AUC on the validation data. Through this, it was confirmed that the grade of glioma was effectively classified by learning the internal relationship between various modalities.

• International Journal of Advanced Culture Technology
• /
• v.10 no.4
• /
• pp.536-546
• /
• 2022

A society will lose a lot of something in this field when the forest fire broke out. If a forest fire can be detected in advance, damage caused by the spread of forest fires can be prevented early. So, we studied how to detect forest fires using CCTV currently installed. In this paper, we present a deep learning-based model through efficient image data construction for monitoring forest fire smoke, which is unstructured data, based on the deep learning model YOLOv5. Through this study, we conducted a study to accurately detect forest fire smoke, one of the amorphous objects of various forms, in YOLOv5. In this paper, we introduce a method of self-learning by producing insufficient data on its own to increase accuracy for unstructured object recognition. The method presented in this paper constructs a dataset with a fixed labelling position for images containing objects that can be extracted from the original image, through the original image and a model that learned from it. In addition, by training the deep learning model, the performance(mAP) was improved, and the errors occurred by detecting objects other than the learning object were reduced, compared to the model in which only the original image was learned.