딥 러닝 기반 이미지 트레이닝을 활용한 하천 공간 내 피복 분류 가능성 검토

강우철,장은경 응용생태공학회 2022 Ecology and resilient infrastructure Vol.9 No.4

This study attempted classification through deep learning-based image training for land cover classification in river spaces which is one of the important data for efficient river management. For this purpose, land cover classification analysis with the RGB image of the target section based on the category classification index of major land cover map was conducted by using the learning outcomes from the result of labeling. In addition, land cover classification of the river spaces was performed by unsupervised and supervised classification from Sentinel-2 satellite images provided in an open format, and this was compared with the results of deep learning-based image classification. As a result of the analysis, it showed more accurate prediction results compared to unsupervised classification results, and it presented significantly improved classification results in the case of high-resolution images. The result of this study showed the possibility of classifying water areas and wetlands in the river spaces, and if additional research is performed in the future, the deep learning based image train method for the land cover classification could be used for river management. 본 연구는 효율적인 하천 관리를 위해 중요한 데이터 중 하나인 하천 공간의 토지피복 분류를 위해 딥 러닝 기반의이미지 트레이닝 방법의 활용가능성을 검토하였다. 이를 위해 대상 구간의 RGB 이미지를 활용하여 라벨링 작업 후 학습시킨결과를 활용하여 기존 대분류 지표를 기준으로 토지피복 분류를 시도하였다. 또한 개방형으로 제공되는 Sentinel-2 위성 영상으로부터무감독 분류 및 감독 분류에 의한 하천 공간의 토지피복 분류를 수행하였으며, 딥 러닝 기반 이미지 분류 결과와 비교하였다. 분석 결과의 경우 무감독 분류 결과와 비교하여 매우 향상된 예측 결과를 보여주었으며, 고해상도 이미지의 경우 더욱 정확한분류 결과를 제시하였다. 단순한 이미지 라벨링을 통해 분류된 피복 분류 결과는 하천 공간 내 수역과 습지의 분류 가능성을보여주었으며, 향후 추가적인 연구 수행이 이루어진다면 하천 관리를 위해 딥 러닝 기반 이미지 트레이닝 기법을 이용한 하천공간내 피복 분류 결과의 활용이 가능할 것으로 판단된다.

Business Application of Convolutional Neural Networks for Apparel Classification Using Runway Image

Yian Seo(서이안),Kyung-shik Shin(신경식) 한국지능정보시스템학회 2018 지능정보연구 Vol.24 No.3

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people’s movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven’t been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image

고해상도 위성영상을 이용한 농촌 도로 매핑을 위한 영상 분류 및 영상 분할 방법 비교에 관한 연구

정윤재 ( Yun-jae Choung ),구본엽 ( Bon-yup Gu ) 한국지리정보학회 2021 한국지리정보학회지 Vol.24 No.3

농촌 도로는 농촌 지역의 개발과 관리를 위한 핵심 기반시설로서 원격탐사 자료를 활용한 농촌도로 관리 기술은 농촌 교통 인프라 확대, 농촌 주민의 삶의 질 개선을 위해 매우 중요하다. 본 연구에서는 농촌 지역을 촬영한 고해상도 위성영상을 활용하여 농촌 도로를 매핑하기 위해 영상 분류 방법과 영상 분할 방법을 다음의 과정을 통하여 비교하였다. 영상 분류의 경우, 심층 신경망 기반 딥러닝 기법을 주어진 고해상도 위성영상에 적용하여 고정밀 객체 분류 지도를 제작하였고 이로부터 농촌 도로 객체를 추출함으로써 농촌 도로를 매핑하였다. 영상 분할의 경우, multiresolution segmentation 기법을 동일한 위성영상에 적용하여 세그먼트 영상을 제작하였고 농촌 도로에 위치한 다중 객체들을 선택하고 이들을 최종적으로 융합하여 농촌 도로를 매핑하였다. 영상 분류 및 영상 분할 방법을 통해 매핑한 농촌 도로의 정확도 검증을 위해 100개의 검사점을 사용하였고 다음과 같은 결론을 도출하였다. 영상 분류 방법에서는 객체 분류 지도 내 오분류 에러로 인해 영상 내 일부 농촌 도로의 인식이 불가능하였으나 영상 분할 방법에서는 영상 내 모든 농촌 도로의 인식이 가능하였으므로 영상 분할 방법이 영상 분류 방법보다 위성영상을 이용한 농촌 도로 매핑 작업에 더 적합한 방법이었다. 그러나 영상 분할 방법을 통해 매핑한 농촌 도로를 구성하는 일부 세그먼트들이 농촌 도로 외 객체를 포함하고 있어 영상 내 일부 농촌 도로에서 오분류 에러가 발생하였다. 추후 연구에서는 객체 기반 분류 또는 합성곱 신경망 등 다양한 정밀 객체 인식 기법을 고해상도 위성영상에 적용하여 농촌 도로의 정확도를 개선할 계획이다. Rural roads are the significant infrastructure for developing and managing the rural areas, hence the utilization of the remote sensing datasets for managing the rural roads is necessary for expanding the rural transportation infrastructure and improving the life quality of the rural residents. In this research, the two different methods such as image classification and image segmentation were compared for mapping the rural road based on the given high-resolution satellite image acquired in the rural areas. In the image classification method, the deep learning with the multiple neural networks was employed to the given high-resolution satellite image for generating the object classification map, then the rural roads were mapped by extracting the road objects from the generated object classification map. In the image segmentation method, the multiresolution segmentation was employed to the same satellite image for generating the segment image, then the rural roads were mapped by merging the road objects located on the rural roads on the satellite image. We used the 100 checkpoints for assessing the accuracy of the two rural roads mapped by the different methods and drew the following conclusions. The image segmentation method had the better performance than the image classification method for mapping the rural roads using the give satellite image, because some of the rural roads mapped by the image classification method were not identified due to the miclassification errors occurred in the object classification map, while all of the rural roads mapped by the image segmentation method were identified. However some of the rural roads mapped by the image segmentation method also had the miclassfication errors due to some rural road segments including the non-rural road objects. In future research the object-oriented classification or the convolutional neural networks widely used for detecting the precise objects from the image sources would be used for improving the accuracy of the rural roads using the high-resolution satellite image.

카테고리 계층을 고려한 회선신경망의 이미지 분류

정노권,조수선 한국멀티미디어학회 2018 멀티미디어학회논문지 Vol.21 No.12

In order to improve the performance of image classifications using Convolutional Neural Networks (CNN), applying a category hierarchy to the classification can be a useful idea. However, the visual separation of object categories is very different according to the upper and lower category levels and highly uneven in image classifications. Therefore, it is doubtable whether the use of category hierarchies for classification is effective in CNN. In this paper, we have clarified whether the image classification using category hierarchies improves classification performance, and found at which level of hierarchy classification is more effective. For experiments we divided the image classification task according to the upper and lower category levels and assigned image data to each CNN model. We identified and compared the results of three classification models and analyzed them. Through the experiments, we could confirm that classification effectiveness was not improved by reduction of number of categories in a classification model. And we found that only with the re-training method in the last network layer, the performance of lower category classification was not improved although that of higher category classification was improved.

A Novel Image Fusion Algorithm Combining with Classification in NCST Domain

Jitao Zhang,Aili Wang,Jiaying Zhao 보안공학연구지원센터 2016 International Journal of Signal Processing, Image Vol.9 No.10

Image fusion is an important branch of information fusion, which is widely used in various fields. At present, the image fusion method is mainly aimed at the different frequency information of the images, the images are fused in transform domain. But in practical application, image fusion is used to improve the credibility of the target information and the demand of background information of is not high. Therefore, this paper puts forward an image fusion method combining with image classification. Firstly, the NSCT transform is used to transform the source images, and the K-Means method is used to realize the classification of the target and the background, and the different fusion criteria are used to get the target and the background. The experimental results show that the image fusion based classification method has a better effect on the subjective visual effect and objective evaluation index.

Image Classification Algorithm Based on Proposal Region Clustering Learning-Unsupervised Deep Learning

Li Lei,Yin Xiao-li 대한전기학회 2023 Journal of Electrical Engineering & Technology Vol.18 No.2

Although deep learning has achieved certain results in image classification, images are susceptible to factors such as lighting conditions, shooting angles, complex backgrounds, rotation transformations or scale scaling, and image data sets in some areas are difficult to obtain. They make the deep learning framework unable to give full play to its generalization ability and nonlinear modeling ability in image classification. Therefore, this paper first proposes a proposal region clustering learning algorithm, which clusters the proposal regions in each image so that each cluster corresponds to the category of the image. Then, different clusters can be regarded as different multi-instance learning packets, and each packet uses the multi-instance learning method to learn the unsupervised image classification detector. It can effectively improve the generalization and modeling capabilities of deep learning models. In addition, this paper proposes an unsupervised deep learning method, which designs an unsupervised deep learning network structure and loss function according to the characteristics of the classified image, and combines densely connected blocks to extract features from the source image. It retains the multi-scale features of the middle layer of the classified image, and effectively solves the problem of insufficient image feature extraction information caused by the lack of image data. It also guarantees the accuracy of subsequent image classification. The experimental results show that the image classification method proposed in this paper not only solves the problem of insufficient image data sets and the interference of various complex factors, but also can accurately classify various image data sets. The accuracy of the image classification method proposed in this paper is 1.38–19% higher than other mainstream deep learning methods.

The Contribution of Feature Selection and Morphological Operation For On-Line Business System’s Image Classification

Mokhairi Makhtar,Nur Shazwani Kamarudin,Syed Abdullah Fadzli,Mumtazimah Mohamad,Fatma Susilawati Mohamad,Mohd Fadzil Abdul Kadir 보안공학연구지원센터 2015 International Journal of Multimedia and Ubiquitous Vol.10 No.11

Automatic image annotation is one of crucial and attractive field of image retrieval. Classification process is part of the important phase in automatic image annotation (AIA). With the explosive growth of methods in this research area, this paper proposes 5 processing steps before image annotation using Amazon dataset, i.e., image segmentation, object identification, feature extraction, feature selection and image features classification. A lot of research has been done in creating numbers of different approaches and algorithm for image segmentation. Otsu is one of the most well known method in image segmentation region based. The proposed model aims to provide the highest accuracy after undergo those processing steps. This paper conducted several experiments for image classification starting from image segmentation in order to demonstrate usefulness and competiveness among different type of classifiers. It also target to study the effect of morphological operation and feature selection to the accuracy. For the classification experiment, it was tested using four types of classifiers: BayesNet, NaiveBayesUpdateable, RandomTree and IBk.

An Efficient Algorithm for Facial Image Classification

Dr.S.Vijayarani,M.Vinupriya 보안공학연구지원센터 2015 International Journal of Signal Processing, Image Vol.8 No.12

Image mining is one of the data mining research areas and it can be defined as getting hidden information from the image databases. It is used to identify unknown patterns, inherent and valuable information from images. Image mining helps to make relationships between various categories of images which are found in large image databases. These images can reveal useful information to the users. Image mining is distinct from low-level computer vision and image processing techniques. It uses methods from computer vision, image retrieval, image processing, data mining, database, machine learning, and artificial intelligence. Although all these subjects study the same object image, the vital difference between image data mining and the other subjects is, image data mining focuses on large scale set of images while image processing and pattern recognition analysis are based on only single image. Face detection is the problem of determining whether a sub-window of an image contains a face. It has received much attention and has been an extensive research topic in recent years. In this research work, facial images are classified based on its shape feature using optimization algorithms. A new algorithm, i.e. classification based similarity finding is proposed for classifying the facial images as round or oval shape. The performance of the proposed classification based similarity algorithm is compared with the particle swam optimization and genetic algorithms. The results of the existing and proposed algorithms are analyzed based on accuracy and execution time factors. From this we observed that the proposed classification based similarity finding algorithm has produced good results.

하이퍼스펙트럴영상 분류에서 정준상관분류기법의 유용성

박민호(Park Min-Ho) 대한토목학회 2006 대한토목학회논문집 D Vol.26 No.5D

본 논문의 의도는 하이퍼스펙트럴 영상의 다량의 밴드를 사용하면서도 효율적인 분류기법의 개발에 초점을 두고 있다. 본 연구에서는 하이퍼스펙트럴 영상의 분류에 있어 이론적으로 밴드수가 많아질수록 분류정확도가 높을 것이라 예상되는, 다변량 통계분석기법중의 하나인 정준상관분석을 적용한 분류기법을 제안한다. 그리고 기존의 대표적인 전통적 분류기법인 최대 우도분류 방법과 비교한다. 사용되는 하이퍼스펙트럴 영상은 2001년 9월 2일 취득된 EO1-Hyperion 영상이다. 실험을 위한 밴드수는 LANDSAT TM 영상에서 열밴드를 제외한 나머지 데이터의 파장대와 일치하는 부분을 감안하여 30개 밴드로 선정하였다. 지상실제데이터로서 비교기본도를 채택하였다. 이 비교기본도와 시각적으로 윤곽을 비교하고, 중첩분석하여 정확도를 평가하였다. 최대우도분류의 경우 수역 분류를 제외하고는 전혀 분류기법으로서의 역할을 하지 못하는 것으로 판단되며, 수역의 경우도 큰 호수 외에 작은 호수나 골프장내 연못, 부분적으로 물이 존재하는 작은 영역 등은 전혀 분류하지 못하고 있는 것으로 나타났다. 그러나 정준상관분류결과는 비교기본도와 형태적으로 시각적 비교를 해볼 때 골프장잔디를 거의 명확히 분류해 내고 있으며, 도시역에 대해서도 고속도로의 선형 등을 상당히 잘 분류해내고 있음을 알 수 있다. 또한 수역의 경우도 골프장 연못이나 대학교내 연못, 기타지역의 연못, 웅덩이 등 까지도 잘 분류해내고 있음을 확인할 수 있다. 결과적으로 정준상관분석 알고리즘의 개념상 트레이닝 영역 선정시 시행착오를 겪지 않고도 정확한 분류를 할 수 있었다. 또한 분류항목 중에서 잔디와 그 외 식물을 구분해 내는 능력과 수역을 추출해 내는 능력이 최대우도분류기법에 비해 우수하였다. 이상의 결과로 판단해 볼 때 하이퍼스펙트럴영상에 적용되는 정준상관분류기법은 농작물 작황 예측과 지표수 탐사에 매우 유용하리라 판단되며, 나아가서는 분광적 고해상도 영상인 하이퍼스펙트럴 데이터를 이용한 GIS 데이터베이스 구축에 중요한 역할을 할 수 있을 것으로 기대된다. The purpose of this study is focused on the development of the effective classification technique using ultra multi band of hyperspectral image. This study suggests the classification technique using canonical correlation analysis, one of multivariate statistical analysis in hyperspectral image classification. High accuracy of classification result is expected for this classification technique as the number of bands increase. This technique is compared with Maximum Likelihood Classification(MLC). The hyperspectral image is the EOl-hyperion image acquired on September 2, 2001, and the number of bands for the experiment were chosen at 30, considering the band scope except the thermal band of Landsat TM. We chose the comparing base map as Ground Truth Data. We evaluate the accuracy by comparing this base map with the classification result image and performing overlay analysis visually. The result showed us that in MLC's case, it can't classify except water, and in case of water, it only classifies big lakes. But Canonical Correlation Classification (CCC) classifies the golf lawn exactly, and it classifies the highway line in the urban area well. In case of water, the ponds that are in golf ground area, the ponds in university, and pools are also classified well. As a result, although the training areas are selected without any trial and error, it was possible to get the exact classification result. Also, the ability to distinguish golf lawn from other vegetations in classification classes, and the ability to classify water was better than MLC technique. Conclusively, this CCC technique for hyperspectral image will be very useful for estimating harvest and detecting surface water. In advance, it will do an important role in the construction of GIS database using the spectral high resolution image, hyperspectral data.

의미적 연관태그와 이미지 내용정보를 이용한 웹 이미지 분류

조수선 ( Soosun Cho ) 한국인터넷정보학회 2010 인터넷정보학회논문지 Vol.11 No.3

본 논문에서는 대용량 온라인 이미지 공유 사이트를 적용 도메인으로 하여 이미지 검색의 만족도를 높이고자 태그의 의미적 연관성과 이미지 자체의 내용 정보를 결합하는 이미지 분류 방법을 제안한다. 이미지 검색 및 분류 알고리즘이 플리커와 같은 대용량 이미지 공유 사이트에서 활용될 수 있으려면 실제 웹상의 태깅된 이미지를 대상으로 한 적용이 가능해야 한다. 제안된 알고리즘은 `bag of visual word`기반의 이미지 내용으로 웹 이미지를 분류하기 위한 것으로서, 의미적 연관태그를 이용해 일차 검색된 이미지들을 훈련 데이터로 사용하여 카테고리 모델을 훈련하고, PLSA를 적용하여 평가 이미지들을 분류하는 것이다. 제안된 방법으로 플리커의 웹 이미지들을 대상으로 실험한 결과, 태그 정보를 이용한 기존의 방법에 비해 우수한 검색 정확도 및 재현율을 확인할 수 있었다. In this paper, we propose an image classification which combines semantic relations of tags with contents of images to improve the satisfaction of image retrieval on application domains as huge image sharing sites. To make good use of image retrieval or classification algorithms on huge image sharing sites as Flickr, they are applicable to real tagged Web images. To classify the Web images by `bag of visual word` based image content, our algorithm includes training the category model by utilizing the preliminary retrieved images with semantically related tags as training data and classifying the test images based on PLSA. In the experimental results on the Flickr Web images, the proposed method produced the better precision and recall rates than those from the existing method using tag information.