• Journal of Korea Multimedia Society
• /
• v.22 no.2
• /
• pp.136-145
• /
• 2019

The crowded environment of a domestic pig farm is highly vulnerable to the spread of infectious diseases such as foot-and-mouth disease, and studies have been conducted to automatically analyze behavior of pigs in a crowded pig farm through a video surveillance system using a camera. Although it is required to correctly separate occluding pigs for tracking each individual pigs, extracting the boundaries of the occluding pigs fast and accurately is a challenging issue due to the complicated occlusion patterns such as X shape and T shape. In this study, we propose a fast and accurate method to separate occluding pigs not only by exploiting the characteristics (i.e., one of the fast deep learning-based object detectors) of You Only Look Once, YOLO, but also by overcoming the limitation (i.e., the bounding box-based object detector) of YOLO with the test-time data augmentation of rotation. Experimental results with two-pigs occlusion patterns show that the proposed method can provide better accuracy and processing speed than one of the state-of-the-art widely used deep learning-based segmentation techniques such as Mask R-CNN (i.e., the performance improvement over Mask R-CNN was about 11 times, in terms of the accuracy/processing speed performance metrics).

• Journal of the Korean Society of Radiology
• /
• v.13 no.1
• /
• pp.111-118
• /
• 2019

Chest digital tomosynthesis has become a practical imaging modality because it can solve the problem of anatomy overlapping in conventional chest radiography. However, because of both limited scan angle and finite-size detector, a portion of chest cannot be represented in some or all of the projection. These bring a discontinuity in intensity across the field of view boundaries in the reconstructed slices, which we refer to as the truncation artifacts. The purpose of this study was to reduce truncation artifacts using a weighted normalization approach and to investigate the performance of this approach for our prototype chest digital tomosynthesis system. The system source-to-image distance was 1100 mm, and the center of rotation of X-ray source was located on 100 mm above the detector surface. After obtaining 41 projection views with ${\pm}20^{\circ}$ degrees, tomosynthesis slices were reconstructed with the filtered back projection algorithm. For quantitative evaluation, peak signal to noise ratio and structure similarity index values were evaluated after reconstructing reference image using simulation, and mean value of specific direction values was evaluated using real data. Simulation results showed that the peak signal to noise ratio and structure similarity index was improved respectively. In the case of the experimental results showed that the effect of artifact in the mean value of specific direction of the reconstructed image was reduced. In conclusion, the weighted normalization method improves the quality of image by reducing truncation artifacts. These results suggested that weighted normalization method could improve the image quality of chest digital tomosynthesis.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.44 no.4
• /
• pp.43-54
• /
• 2007

This paper presents design and verification of a face detection hardware for real time application. Face detection algorithm detects rough face position based on already acquired feature parameter data. The hardware is composed of five main modules: Integral Image Calculator, Feature Coordinate Calculator, Feature Difference Calculator, Cascade Calculator, and Window Detection. It also includes on-chip Integral Image memory and Feature Parameter memory. The face detection hardware was verified by using S3C2440A CPU of Samsung Electronics, Virtex4LX100 FPGA of Xilinx, and a CCD Camera module. Our design uses 3,251 LUTs of Xilinx FPGA and takes about 1.96${\sim}$0.13 sec for face detection depending on sliding-window step size, when synthesized for Virtex4LX100 FPGA. When synthesized on Magnachip 0.25um ASIC library, it uses about 410,000 gates (Combinational area about 345,000 gates, Noncombinational area about 65,000 gates) and takes less than 0.5 sec for face realtime detection. This size and performance shows that it is adequate to use for embedded system applications. It has been fabricated as a real chip as a part of XF1201 chip and proven to work.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.7
• /
• pp.1235-1242
• /
• 2008

In this paper, x-ray image sensor of photon counting type having a $32{\times}32$ pixel array is designed with $0.18{\mu}m$ triple-well CMOS process. Each pixel of the designed image sensor has an area of loot $100{\times}100\;{\mu}m2$ and is composed of about 400 transistors. It has an open pad of an area of $50{\times}50{\mu}m2$ of CSA(charge Sensitive Amplifier) with x-ray detector through a bump bonding. To reduce layout size, self-biased folded cascode CMOS OP amp is used instead of folded cascode OP amp with voltage bias circuit at each single-pixel CSA, and 15-bit LFSR(Linear Feedback Shift Register) counter clock generator is proposed to remove short pulse which occurs from the clock before and after it enters the counting mode. And it is designed that sensor data can be read out of the sensor column by column using a column address decoder to reduce the maximum current of the CMOS x-ray image sensor in the readout mode.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.11
• /
• pp.457-464
• /
• 2019

The weight of pig is one of the main factors in determining the health and growth state of pigs, their shipment, the breeding environment, and the ration of feed, and thus measuring the pig's weight is an important issue in productivity perspective. In order to estimate the pig's weight by using the number of pig's pixels from images, acquired from a Top-view camera, the posture determining and the head removal from images are necessary to measure the accurate number of pixels. In this research, we propose the fast and accurate method to determine the pig's posture by using a fast image processing technique, find the head location by using a fast deep learning technique, and remove pig's head by using light weighted image processing technique. First, we determine the pig's posture by comparing the length from the center of the pig's body to the outline of the pig in the binary image. Then, we train the location of pig's head, body, and hip in images using YOLO(one of the fast deep learning based object detector), and then we obtain the location of pig's head and remove an outside area of head by using head location. Finally, we find the boundary of head and body by using Convex-hull, and we remove pig's head. In the Experiment result, we confirmed that the pig's posture was determined with an accuracy of 0.98 and a processing speed of 250.00fps, and the pig's head was removed with an accuracy of 0.96 and a processing speed of 48.97fps.

• Journal of the Korean Society of Radiology
• /
• v.17 no.1
• /
• pp.1-9
• /
• 2023

For optimal image quality of computer tomography angiography (CTA), different iodine concentrations and scan parameters were applied to quantitatively evaluate the image quality characteristics of filtered back projection (FBP), hybrid-iterative reconstruction (hybrid-IR), and deep learning reconstruction (DLR). A 320-row-detector CT scanner scanned a phantom with various iodine concentrations (1.2, 2.9, 4.9, 6.9, 10.4, 14.3, 18.4, and 25.9 mg/mL) located at the edge of a cylindrical water phantom with a diameter of 19 cm. Data obtained using each reconstruction technique was analyzed through noise, coefficient of variation (COV), and root mean square error (RMSE). As the iodine concentration increased, the CT number value increased, but the noise change did not show any special characteristics. COV decreased with increasing iodine concentration for FBP, adaptive iterative dose reduction (AIDR) 3D, and advanced intelligent clear-IQ engine (AiCE) at various tube voltages and tube currents. In addition, when the iodine concentration was low, there was a slight difference in COV between the reconstitution techniques, but there was little difference as the iodine concentration increased. AiCE showed the characteristic that RMSE decreased as the iodine concentration increased but rather increased after a specific concentration (4.9 mg/mL). Therefore, the user will have to consider the characteristics of scan parameters such as tube current and tube voltage as well as iodine concentration according to the reconstruction technique for optimal CTA image acquisition.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1057-1068
• /
• 2022

This paper compares the combination performance of feature point-based matching algorithms as a study to confirm the matching possibility between image taken by a user and a virtual texture image with the goal of developing mobile-based real-time image positioning technology. The feature based matching algorithm includes process of extracting features, calculating descriptors, matching features from both images, and finally eliminating mismatched features. At this time, for matching algorithm combination, we combined the process of extracting features and the process of calculating descriptors in the same or different matching algorithm respectively. V-World 3D desktop was used for the virtual indoor texture image. Currently, V-World 3D desktop is reinforced with details such as vertical and horizontal protrusions and dents. In addition, levels with real image textures. Using this, we constructed dataset with virtual indoor texture data as a reference image, and real image shooting at the same location as a target image. After constructing dataset, matching success rate and matching processing time were measured, and based on this, matching algorithm combination was determined for matching real image with virtual image. In this study, based on the characteristics of each matching technique, the matching algorithm was combined and applied to the constructed dataset to confirm the applicability, and performance comparison was also performed when the rotation was additionally considered. As a result of study, it was confirmed that the combination of Scale Invariant Feature Transform (SIFT)'s feature and descriptor detection had the highest matching success rate, but matching processing time was longest. And in the case of Features from Accelerated Segment Test (FAST)'s feature detector and Oriented FAST and Rotated BRIEF (ORB)'s descriptor calculation, the matching success rate was similar to that of SIFT-SIFT combination, while matching processing time was short. Furthermore, in case of FAST-ORB, it was confirmed that the matching performance was superior even when 10° rotation was applied to the dataset. Therefore, it was confirmed that the matching algorithm of FAST-ORB combination could be suitable for matching between virtual texture image and real image.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.2
• /
• pp.33-44
• /
• 2010

This paper propose the structure of real-time face detection hardware architecture for robot vision processing applications. The proposed architecture is robust against illumination changes and operates at no less than 60 frames per second. It uses Modified Census Transform to obtain face characteristics robust against illumination changes. And the AdaBoost algorithm is adopted to learn and generate the characteristics of the face data, and finally detected the face using this data. This paper describes the face detection hardware structure composed of Memory Interface, Image Scaler, MCT Generator, Candidate Detector, Confidence Comparator, Position Resizer, Data Grouper, and Detected Result Display, and verification Result of Hardware Implementation with using Virtex5 LX330 FPGA of Xilinx. Verification result with using the images from a camera showed that maximum 32 faces per one frame can be detected at the speed of maximum 149 frame per second.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_2
• /
• pp.737-746
• /
• 2022

Utilizing weather independent SAR images along with machine learning based object detector is effective in robust vessel monitoring. While conventional SAR images often applied amplitude data from Single Look Complex, exploitation of polarimetric parameters acquired from multiple polarimetric SAR images was yet to be implemented to vessel detection utilizing machine learning. Hence, this study used four polarimetric parameters (H, p₁, DoP, DPRVI) retrieved from eigen-decomposition and two backscattering coefficients (γ₀, _VV, γ₀, _VH) from radiometric calibration; six bands in total were respectively exploited from 52 Sentinel-1 SAR images, accompanied by vessel training data extracted from AIS information which corresponds to acquisition time span of the SAR image. Evaluating different cases of combination, the use of polarimetric indexes along with amplitude values derived enhanced vessel detection performances than that of utilizing amplitude values exclusively.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.15 no.1
• /
• pp.58-64
• /
• 2011

Objective: The 3-dimensional reconstruction method with resolution recovery modeling has advantages of high spatial resolution and contrast because of its precise modeling of spatial blurring according to the distance from detector plane. The aim of this study was to evaluate one of the resolution recovery reconstruction methods (Astonish, Philips Medical), compare it to other iterative reconstructions, and verify its clinical usefulness. Materials and Methods: NEMA IEC PET body phantom and Flanges Jaszczak ECT phantom (Data Spectrum Corp., USA) studies were performed using Skylight SPECT (Philips) system under four different conditions; short or long (2 times of short) radius, and half or full (40 kcts/frame) acquisition counts. Astonish reconstruction method was compared with two other iterative reconstructions; MLEM and 3D-OSEM which vendor supplied. For quantitative analysis, the contrast ratios obtained from IEC phantom test were compared. Reconstruction parameters were determined by optimization study using graph of contrast ratio versus background variability. The qualitative comparison was performed with Jaszczak ECT phantom and human myocardial data. Results: The overall contrast ratio was higher with Astonish than the others. For the largest hot sphere of 37 mm diameter, Astonish showed about 27.1% and 17.4% higher contrast ratio than MLEM and 3D-OSEM, in short radius study. For long radius, Astonish showed about 40.5% and 32.6% higher contrast ratio than MLEM and 3D-OSEM. The effect of acquired counts was insignificant. In the qualitative studies with Jaszczak phantom and human myocardial data, Astonish showed the best image quality. Conclusion: In this study, we have found out that Astonish can provide more reliable clinical results by better image quality compared to other iterative reconstruction methods. Although further clinical studies are required, Astonish would be used in clinics with confidence for enhancement of images.