• International Journal of Contents
• /
• v.15 no.4
• /
• pp.8-15
• /
• 2019

This paper presents a solution of the 'Quick, Draw! Doodle Recognition Challenge' hosted by Google. Doodles are drawings comprised of concrete representational meaning or abstract lines creatively expressed by individuals. In this challenge, a doodle is presented as a sequence of sketches. From the view of at the sketch level, to learn the pattern of strokes representing a doodle, we propose a sequential model stacked with multiple convolution layers and Long Short-Term Memory (LSTM) cells following the attention mechanism [15]. From the view at the image level, we use multiple models pre-trained on ImageNet to recognize the doodle. Finally, an ensemble and a post-processing method using the minimum cost flow algorithm are introduced to combine multiple models in achieving better results. In this challenge, our solutions garnered 11th place among 1,316 teams. Our performance was 0.95037 MAP@3, only 0.4% lower than the winner. It demonstrates that our method is very competitive. The source code for this competition is published at: https://github.com/ngxbac/Kaggle-QuickDraw.

• Journal of Forest and Environmental Science
• /
• v.29 no.2
• /
• pp.131-137
• /
• 2013

This study attempts to classify forest species using hyperspectral data for supporting resources management. The primary dataset used was AISA sensor. The sensor was mounted onboard the NOMAD GAF-27 aircraft at 2,000 m altitude creating a 2 m spatial resolution on the ground. Pre-processing was carried out with CALIGEO software, which automatically corrects for both geometric and radiometric distortions of the raw image data. The radiance data set was then converted to at-sensor reflectance derived from the FODIS sensor. Spectral Angle Mapper (SAM) technique was used for image classification. The spectra libraries for tree species were established after confirming the appropriate match between field spectra and pixel spectra. Results showed that the highest spectral signature in NIR range were Kembang Semangkok (Scaphium macropodum), followed by Meranti Sarang Punai (Shorea parvifolia) and Chengal (Neobalanocarpus hemii). Meanwhile, the lowest spectral response were Kasai (Pometia pinnata), Kelat (Eugenia spp.) and Merawan (Hopea beccariana), respectively. The overall accuracy obtained was 79%. Although the accuracy of SAM techniques is below the expectation level, SAM classifier was able to classify tropical tree species. In future it is believe that the most effective way of ground data collection is to use the ground object that has the strongest response to sensor for more significant tree signatures.

• Smart Media Journal
• /
• v.7 no.2
• /
• pp.23-33
• /
• 2018

A musical sheet is read by optical music recognition (OMR) systems that automatically recognize and reconstruct the read data to convert them into a machine-readable format such as XML so that the music can be played. This process, however, is very challenging due to the large variety of musical styles, symbol notation, and other distortions. In this paper, we present a model for the recognition of musical symbols through the use of a mobile application, whereby a camera is used to capture the input image; therefore, additional difficulties arise due to variations of the illumination and distortions. For our proposed model, we first generate a line adjacency graph (LAG) to remove the staff lines and to perform primitive detection. After symbol segmentation using the primitive information, we use a covariance-matching method to estimate the similarity between every symbol and pre-defined templates. This method generates the three hypotheses with the highest scores for likelihood measurement. We also add a global consistency (time measurements) to verify the three hypotheses in accordance with the structure of the musical sheets; one of the three hypotheses is chosen through a final decision. The results of the experiment show that our proposed method leads to promising results.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.2
• /
• pp.433-438
• /
• 2019

This paper proposes a method of improving deep learning based numbers and characters recognition performance on parts of drawing through image preprocessing. The proposed character recognition system consists of image preprocessing and 7 layer deep learning model. Mathematical morphological filtering is used as preprocessing to remove the lines and shapes which causes false recognition of numbers and characters on parts drawing. Further.. Further, the used deep learning model is a 7 layer deep learning model instead of VGG-16 model. As a result of the proposed OCR method, the recognition rate of characters is 92.57% and the precision is 92.82%.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.2 s.314
• /
• pp.116-125
• /
• 2007

In this paper, we propose a illumination environment adaptive real-time surveillance system for security of important area such as military bases, prisons, and strategic infra structures. The proposed system recognizes movement of objects on the bright environments as well as in dark illumination. The procedure of proposed system may be summarized as follows. First, the system discriminates between bright and dark with input image distribution. Then, if the input image is dark, the system has a pre-processing. The Multi-scale Retinex Color Restoration(MSRCR) is processed to enhance the contrast of image captured in dark environments. Secondly, the enhanced input image is subtracted with the revised background image. And then, we take a morphology image processing to obtain objects correctly. Finally, each bounding box enclosing each objects are tracked. The center point of each bounding box obtained by the proposed algorithm provides more accurate tracking information. Experimental results show that the proposed system provides good performance even though an object moves very fast and the background is quite dark.

• The KIPS Transactions:PartB
• /
• v.13B no.1 s.104
• /
• pp.53-62
• /
• 2006

Object recognition techniques using principal component analysis are disposed to be decreased recognition rate when lighting change of image happens. The purpose of this thesis is to propose an object recognition technique using new PCA analysis method that discriminates an object in database even in the case that the variation of illumination in training images exists. And the object recognition algorithm proposed here represents more enhanced recognition rate using improved k-Nearest Neighbor. In this thesis, we proposed an object recognition algorithm which creates object space by pre-processing and being learned image using histogram equalization and median filter. By spreading histogram of test image using histogram equalization, the effect to change of illumination is reduced. This method is stronger to change of illumination than basic PCA method and normalization, and almost removes effect of illumination, therefore almost maintains constant good recognition rate. And, it compares ingredient projected test image into object space with distance of representative value and recognizes after representative value of each object in model image is made. Each model images is used in recognition unit about some continual input image using improved k-Nearest Neighbor in this thesis because existing method have many errors about distance calculation.

• Korean Journal of Remote Sensing
• /
• v.27 no.4
• /
• pp.495-506
• /
• 2011

COMS In-Orbit Tests(IOT), performed from July, 2010 to Jan, 2011, were successfully completed and the scientific data from MI and GOCI has been distributed officially from April, 2011. This paper focuses on the geometric calibration system tests conducted during the IOT. The geometric calibration process, which is one of the primary objectives of the IOT is the final step of COMS data pre-processing. The basic principles of the geometric calibration (or image navigation and registration, INR) algorithm for COMS are described and the functional and performance tests of COMS INR system were summarized according to the COMS IOT phases. Final performance testes were carried out using data sets acquired from the real-time COMS data pre-processing system. Geometric calibration accuracy of the COMS data showed excellent quality and met requirement specifications.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.5
• /
• pp.1155-1163
• /
• 2011

In this paper, we propose a new pre-processing algorithm applied to depth map to improve the coding efficiency. Now, 3DV/FTV group in the MPEG is working for standard of 3DVC(3D video coding), but compression method for depth map images are not confirmed yet. In the proposed algorithm, after dividing the histogram distribution of a given depth map by EM clustering method based on GMM, we classify the depth map into several layered images. Then, we apply different mean shift filter to each classified image according to the existence of background or foreground in it. In other words, we try to maximize the coding efficiency while keeping the boundary of each object and taking average operation toward inner field of the boundary. The experiments are performed with many test images and the results show that the proposed algorithm achieves bits reduction of 19% ~ 20% and computation time is also reduced.

• Journal of Korea Society of Industrial Information Systems
• /
• v.27 no.1
• /
• pp.19-26
• /
• 2022

This paper presents a pre-processing method to facilitate the container's BIC-code recognition. We propose a network that can find ROI(Region Of Interests) containing a BIC-code region and estimate a homography matrix for warping. Taking the structure of STN(Spatial Transformer Networks), the proposed network consists of next 3 steps, ROI detection, homography matrix estimation, and warping using the homography estimated in the previous step. It contributes to improving the accuracy of BIC-code recognition by estimating ROI and matrix using the proposed network and correcting perspective distortion of ROI using the estimated matrix. For performance evaluation, five evaluators evaluated the output image as a perfect score of 5 and received an average of 4.25 points, and when visually checked, 224 out of 312 photos are accurately and perfectly corrected, containing ROI.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.12C
• /
• pp.1029-1038
• /
• 2008

In this paper, we propose a method for an arbitrary-view image generation in multi-view video and methods for pre- and post-processing to compensate unattended regions in the generated image. To generate an arbitrary-view image, camera geometry is used. Three dimensional coordinates of image pixels can be obtained by using depth information of multi-view video and parameter information of multi-view cameras, and by replacing three dimensional coordinates on a two dimensional image plane of other view, arbitrary-view image can be reconstructed. However, the generated arbitrary-view image contains many unattended regions. In this paper, we also proposed a method for compensating these regions considering temporal redundancy and spatial direction of an image and an error of acquired multi-view image and depth information. Test results show that we could obtain a reliably synthesized view-image with objective measurement of PSNR more than 30dB and subjective estimation of DSCQS(double stimulus continuous quality scale method) more than 3.5 point.