• The Journal of the Convergence on Culture Technology
• /
• v.10 no.3
• /
• pp.909-914
• /
• 2024

This study suggests a quantitative evaluation of transfer learning, which is widely used in various AI fields, including image recognition for robot vision. Quantitative and qualitative analyses of results applying transfer learning are presented, but transfer learning itself is not discussed. Therefore, this study proposes a quantitative evaluation of transfer learning itself based on MNIST, a handwritten digit database. For the reference network, the change in recognition accuracy according to the depth of the transfer learning frozen layer and the ratio of transfer learning data and pre-training data is tracked. It is observed that when freezing up to the first layer and the ratio of transfer learning data is more than 3%, the recognition accuracy of more than 90% can be stably maintained. The transfer learning quantitative evaluation method of this study can be used to implement transfer learning optimized according to the network structure and type of data in the future, and will expand the scope of the use of robot vision and image analysis AI in various environments.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.4
• /
• pp.318-324
• /
• 2017

We introduced a stereo camera on the aircraft to detect flight objects and to estimate the 3D position of them. The Saliency map algorithm based on PCT was proposed to detect a small object between clouds, and then we processed a stereo matching algorithm to find out the disparity between the left and right camera. In order to extract accurate disparity, cost aggregation region was used as a variable region to adapt to detection object. In this paper, we use the detection result as the cost aggregation region. In order to extract more precise disparity, sub-pixel interpolation is used to extract float type-disparity at sub-pixel level. We also proposed a method to estimate the spatial position of an object by using camera parameters. It is expected that it can be applied to image - based object detection and collision avoidance system of autonomous aircraft in the future.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.30 no.3
• /
• pp.249-254
• /
• 2017

The paper presents a new short-term dynamic displacement estimation method based on an acceleration and a geophone sensor. The proposed method combines acceleration and velocity measurements through a real time data fusion algorithm based on Kalman filter. The proposed method can estimate the displacement of a structure without displacement sensors, which is typically difficult to be applied to earthquake or fire sites due to their requirement of a fixed rigid support. The proposed method double-integrates the acceleration measurement recursively, and corrects an accumulated integration error based on the velocity measurement, The performance of the proposed method was verified by a lab-scale test, in which displacement estimated by the proposed method are compared to a reference displacement measured by laser doppler vibrometer (LDV).

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.2
• /
• pp.133-140
• /
• 2017

In this paper, we has proposed an automatic vehicle monitoring system based on computer vision technology. The real-time display system has displayed a system that can be performed in automatic monitoring and control while meeting the essential requirements of ITS. Another advantage has that for a powerful vehicle tracking, the main obstacle handing system, which has the shadow tracking of moving objects. In order to obtain all kinds of information from the tracked vehicle image, the vehicle must be clearly displayed on the surveillance screen. Over time, it's necessary to precisely control the vehicle, and a three-dimensional model-based approach has been also necessary. In general, each type of vehicle has represented by the skeleton of the object or wire frame model, and the trajectory of the vehicle can be measured with high precision in a 3D-based manner even if the system has not running in real time. In this paper, we has applied on segmentation method to vehicle, background, and shadow. The validity of the low level vehicle control tracker was also detected through speed tracking of the speeding car. In conclusion, we intended to improve the improved tracking method in the tracking control system and to develop the highway monitoring and control system.

• The Journal of the Korea Contents Association
• /
• v.19 no.2
• /
• pp.1-10
• /
• 2019

The present research aims to identify the issues that occurred when uploading stock photos to the internet-based stock image agencies and to develop technical solutions based on web service technologies. We identify the issues by examination of previous studies and stock photo uploading systems of major three agencies currently in service. As such, we develop web service technology by focusing on the following matters. First, we apply an automatic tag system to ensure convenience. Second, to ensure safety, we apply a technology that easily enables prevention of portrait rights violations and trademark infringements. We also prepare for measures against possible harmfulness. Third, to ensure completeness, we apply a method which resolves upload failure issues that frequently occurred in the past. In particular, the present research is significant as it applies an automatic image analysis system based on Google Cloud Vision API as the artificial intelligence-based image processing technology. In addition, we develop a web service program which improves user access by using SNS-type screen composition.

• ETRI Journal
• /
• v.36 no.3
• /
• pp.439-449
• /
• 2014

In this paper, a novel vision-based nighttime vehicle detection approach is presented, combining state machines and downhill simplex optimization. In the proposed approach, vehicle detection is modeled as a sequential state transition problem; that is, vehicle arrival, moving, and departure at a chosen detection area. More specifically, the number of bright pixels and their differences, in a chosen area of interest, are calculated and fed into the proposed state machine to detect vehicles. After a vehicle is detected, the location of the headlights is determined using the downhill simplex method. In the proposed optimization process, various headlights were evaluated for possible headlight positions on the detected vehicles; allowing for an optimal headlight position to be located. Simulation results were provided to show the robustness of the proposed approach for nighttime vehicle and headlight detection.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.56 no.1
• /
• pp.55-60
• /
• 2020

In this study, the authors introduce a newly developed flatfish grading system. Owing to the features of flatfish with and wide body, the general types of grading system are not easy to apply for it. Furthermore, the flatfish to be graded is alive such that the existing measurement and grading systems cannot be used for it as well. This study gives a solution for measuring and grading the flatfish with high speed and good accuracy. For this object, the authors developed flatfish measurement and grading system. This system consist of the feeding, conveying, measurement part and sorting part. Especially, the measurement part is made by vision based measuring technique which satisfies the given specification. The result from the experiment shows that the developed system is applicable for measuring and grading the flatfish sizes in variety.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4429-4447
• /
• 2018

Local binary descriptors are well-suited for many real-time and/or large-scale computer vision applications, while their low computational complexity is usually accompanied by the limitation of performance. In this paper, we propose a new optimization framework, RLDB (Robust-LDB), to improve a typical region-based binary descriptor LDB (local difference binary) and maintain its computational simplicity. RLDB extends the multi-feature strategy of LDB and applies a more complete region-comparing configuration. A cascade bit selection method is utilized to select the more representative patterns from massive comparison pairs and an online learning strategy further optimizes descriptor for each specific patch separately. They both incorporate LDP (linear discriminant projections) principle to jointly guarantee the robustness and distinctiveness of the features from various scales. Experimental results demonstrate that this integrated learning framework significantly enhances LDB. The improved descriptor achieves a performance comparable to floating-point descriptors on many benchmarks and retains a high computing speed similar to most binary descriptors, which better satisfies the demands of applications.

• Journal of the Optical Society of Korea
• /
• v.18 no.5
• /
• pp.507-516
• /
• 2014

We present a hybrid method for spectral reflectivity recovery, using 3D extrapolation as a supplemental method for 3D interpolation. The proposed 3D extrapolation is an extended version of 3D interpolation based on the barycentric algorithm. It is faster and more accurate than the conventional spectral-recovery techniques of principal-component analysis and nonnegative matrix transformation. Four different extrapolation techniques (based on nearest neighbors, circumcenters, in-centers, and centroids) are formulated and applied to recover spectral reflectivity. Under the standard conditions of a D65 illuminant and 1964 $10^{\circ}$ observer, all reflectivity data from 1269 Munsell color chips are successfully reconstructed. The superiority of the proposed method is demonstrated using statistical data to compare coefficients of correlation and determination. The proposed hybrid method can be applied for fast and accurate spectral reflectivity recovery in image processing.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.12 no.2
• /
• pp.309-314
• /
• 2017

Recently, the fields such as the service robot, the autonomous driving electric car, and the torpedo ladle cars operated autonomously to enhance the efficiency of management of the steel mill are receiving great attention. But development of automatic power supply that doesn't need human intervention be a problem. In this paper, a position tracking technology of power connector receptacle based on the computer vision is studied which can recognize and identify the position of the power connector receptacle, and finally its possibility is verified using OpenCV program.