• The Journal of the Acoustical Society of Korea
• /
• v.43 no.3
• /
• pp.351-360
• /
• 2024

With the increasing use of drone technology, the Unmanned Aerial Vehicle (UAV) is now being utilized in various fields. However, this increased use of drones has resulted in various issues. Due to its small size, the drone is difficult to detect with radar or optical equipment, so acoustical tracking methods have been recently applied. In this paper, a method of localization of multiple drones using the acoustic characteristics of the quadcopter drone is suggested. Because the acoustic characteristics induced by each rotor are differentiated depending on the type of drone and its movement state, the sound source of the drone can be reconstructed by spatially clustering the results of the estimated positions of the blade passing frequency and its harmonic sound source. The reconstructed sound sources are utilized to finally determine the location of multiple-drone sound sources by applying the source localization algorithm. An experiment is conducted to analyze the acoustic characteristics of the test quadcopter drones, and the simulations for three different types of drones are conducted to localize the multiple drones based on the measured acoustic signals. The test result shows that the location of multiple drones can be estimated by utilizing the acoustic characteristics of the drone. Also, one can see that the clarity of the separated drone sound source and the source localization algorithm affect the accuracy of the localization for multiple-drone sound sources.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.13 no.4
• /
• pp.111-124
• /
• 2010

To monitor the environment of land surface change is considered as an important research field since those parameters are related with land use, climate change, meteorological study, agriculture modulation, surface energy balance, and surface environment system. For the change detection, many different methods have been presented for distributing more detailed information with various tools from ground based measurement to satellite multi-spectral sensor. Recently, using high resolution satellite data is considered the most efficient way to monitor extensive land environmental system especially for higher spatial and temporal resolution. In this study, we use two different spatial resolution satellites; the one is SPOT/VEGETATION with 1 km spatial resolution to detect coarse resolution of the area change and determine objective threshold. The other is Landsat satellite having high resolution to figure out detailed land environmental change. According to their spatial resolution, they show different observation characteristics such as repeat cycle, and the global coverage. By correlating two kinds of satellites, we can detect land surface change from mid resolution to high resolution. The K-mean clustering algorithm is applied to detect changed area with two different temporal images. When using solar spectral band, there are complicate surface reflectance scattering characteristics which make surface change detection difficult. That effect would be leading serious problems when interpreting surface characteristics. For example, in spite of constant their own surface reflectance value, it could be changed according to solar, and sensor relative observation location. To reduce those affects, in this study, long-term Normalized Difference Vegetation Index (NDVI) with solar spectral channels performed for atmospheric and bi-directional correction from SPOT/VEGETATION data are utilized to offer objective threshold value for detecting land surface change, since that NDVI has less sensitivity for solar geometry than solar channel. The surface change detection based on long-term NDVI shows improved results than when only using Landsat.

• Journal of Internet Computing and Services
• /
• v.25 no.1
• /
• pp.99-107
• /
• 2024

Along with economic growth and industrial development, there is an increasing demand for various electronic components and device production of semiconductor, SMT component, and electrical battery products. However, these products may contain foreign substances coming from manufacturing process such as iron, aluminum, plastic and so on, which could lead to serious problems or malfunctioning of the product, and fire on the electric vehicle. To solve these problems, it is necessary to determine whether there are foreign materials inside the product, and may tests have been done by means of non-destructive testing methodology such as ultrasound ot X-ray. Nevertheless, there are technical challenges and limitation in acquiring X-ray images and determining the presence of foreign materials. In particular Small-sized or low-density foreign materials may not be visible even when X-ray equipment is used, and noise can also make it difficult to detect foreign objects. Moreover, in order to meet the manufacturing speed requirement, the x-ray acquisition time should be reduced, which can result in the very low signal- to-noise ratio(SNR) lowering the foreign material detection accuracy. Therefore, in this paper, we propose a five-step approach to overcome the limitations of low resolution, which make it challenging to detect foreign substances. Firstly, global contrast of X-ray images are increased through histogram stretching methodology. Second, to strengthen the high frequency signal and local contrast, we applied local contrast enhancement technique. Third, to improve the edge clearness, Unsharp masking is applied to enhance edges, making objects more visible. Forth, the super-resolution method of the Residual Dense Block (RDB) is used for noise reduction and image enhancement. Last, the Yolov5 algorithm is employed to train and detect foreign objects after learning. Using the proposed method in this study, experimental results show an improvement of more than 10% in performance metrics such as precision compared to low-density images.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.22 no.3
• /
• pp.82-98
• /
• 2019

The importance of spatial information is rapidly rising. In particular, 3D spatial information construction and modeling for Real World Objects, such as smart cities and digital twins, has become an important core technology. The constructed 3D spatial information is used in various fields such as land management, landscape analysis, environment and welfare service. Three-dimensional modeling with image has the hig visibility and reality of objects by generating texturing. However, some texturing might have occlusion area inevitably generated due to physical deposits such as roadside trees, adjacent objects, vehicles, banners, etc. at the time of acquiring image Such occlusion area is a major cause of the deterioration of reality and accuracy of the constructed 3D modeling. Various studies have been conducted to solve the occlusion area. Recently the researches of deep learning algorithm have been conducted for detecting and resolving the occlusion area. For deep learning algorithm, sufficient training data is required, and the collected training data quality directly affects the performance and the result of the deep learning. Therefore, this study analyzed the ability of detecting the occlusion area of the image using various image quality to verify the performance and the result of deep learning according to the quality of the learning data. An image containing an object that causes occlusion is generated for each artificial and quantified image quality and applied to the implemented deep learning algorithm. The study found that the image quality for adjusting brightness was lower at 0.56 detection ratio for brighter images and that the image quality for pixel size and artificial noise control decreased rapidly from images adjusted from the main image to the middle level. In the F-measure performance evaluation method, the change in noise-controlled image resolution was the highest at 0.53 points. The ability to detect occlusion zones by image quality will be used as a valuable criterion for actual application of deep learning in the future. In the acquiring image, it is expected to contribute a lot to the practical application of deep learning by providing a certain level of image acquisition.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_1
• /
• pp.637-654
• /
• 2023

In recent years, the number of users has been increasing with the rapid development of earth observation satellites. In response, the Committee on Earth Observation Satellites (CEOS) has been striving to provide user-friendly satellite images by introducing the concept of Analysis Ready Data (ARD) and defining its requirements as CEOS ARD for Land (CARD4L). In ARD, a mask called an Unusable Data Mask (UDM), identifying unnecessary pixels for land analysis, should be provided with a satellite image. UDMs include clouds, cloud shadows, terrain shadows, etc. Terrain shadows are generated in mountainous terrain with large terrain relief, and these areas cause errors in analysis due to their low radiation intensity. previous research on terrain shadow detection focused on detecting terrain shadow pixels to correct terrain shadows. However, this should be replaced by the terrain correction method. Therefore, there is a need to expand the purpose of terrain shadow detection. In this study, to utilize CAS500-4 for forest and agriculture analysis, we extended the scope of the terrain shadow detection to shaded areas. This paper aims to analyze the potential for terrain shadow detection to make a terrain shadow mask for South and North Korea. To detect terrain shadows, we used a Hill-shade algorithm that utilizes the position of the sun and a surface's derivatives, such as slope and aspect. Using RapidEye images with a spatial resolution of 5 meters and Sentinel-2 images with a spatial resolution of 10 meters over the Korean Peninsula, the optimal threshold for shadow determination was confirmed by comparing them with the ground truth. The optimal threshold was used to perform terrain shadow detection, and the results were analyzed. As a qualitative result, it was confirmed that the shape was similar to the ground truth as a whole. In addition, it was confirmed that most of the F1 scores were between 0.8 and 0.94 for all images tested. Based on the results of this study, it was confirmed that automatic terrain shadow detection was well performed throughout the Korean Peninsula.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.1
• /
• pp.96-110
• /
• 2017

Most of vehicle detection studies using the existing general lens or wide-angle lens have a blind spot in the rear detection situation, the image is vulnerable to noise and a variety of external environments. In this paper, we propose a method that is detection in harsh external environment with noise, blind spots, etc. First, using a fish-eye lens will help minimize blind spots compared to the wide-angle lens. When angle of the lens is growing because nonlinear radial distortion also increase, calibration was used after initializing and optimizing the distortion constant in order to ensure accuracy. In addition, the original image was analyzed along with calibration to remove fog and calibrate brightness and thereby enable detection even when visibility is obstructed due to light and dark adaptations from foggy situations or sudden changes in illumination. Fog removal generally takes a considerably significant amount of time to calculate. Thus in order to reduce the calculation time, remove the fog used the major fog removal algorithm Dark Channel Prior. While Gamma Correction was used to calibrate brightness, a brightness and contrast evaluation was conducted on the image in order to determine the Gamma Value needed for correction. The evaluation used only a part instead of the entirety of the image in order to reduce the time allotted to calculation. When the brightness and contrast values were calculated, those values were used to decided Gamma value and to correct the entire image. The brightness correction and fog removal were processed in parallel, and the images were registered as a single image to minimize the calculation time needed for all the processes. Then the feature extraction method HOG was used to detect the vehicle in the corrected image. As a result, it took 0.064 seconds per frame to detect the vehicle using image correction as proposed herein, which showed a 7.5% improvement in detection rate compared to the existing vehicle detection method.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.12
• /
• pp.51-57
• /
• 2019

In this paper, we analyzed the performance of IoT based smart wearable mine detection device. There are various mine detection methods currently used by the military. Still, in the general field, mine detection is performed by visual detection, probe detection, detector detection, and other detection methods. The detection method by the detector is using a GPR sensor on the detector, which is possible to detect metals, but it is difficult to identify non-metals. It is hard to distinguish whether the area where the detection was performed or not. Also, there is a problem that a lot of human resources and time are wasted, and if the user does not move the sensor at a constant speed or moves too fast, it is difficult to detect landmines accurately. Therefore, we studied the smart wearable mine detection device composed of human body antenna, main microprocessor, smart glasses, body-mounted LCD monitor, wireless data transmission, belt type power supply, black box camera, which is to improve the problem of the error of mine detection using unidirectional ultrasonic sensing signal. Based on the results of this study, we will conduct an experiment to confirm the possibility of detecting underground mines based on the Internet of Things (IoT). This paper consists of an introduction, experimental environment composition, simulation analysis, and conclusion. Introduction introduces the research contents such as mines, mine detectors, and research progress. It consists of large anti-personnel mine, M16A1 fragmented anti-mine, M15 and M19 antitank mines, plastic bottles similar to mines and aluminum cans. Simulation analysis is conducted by using MATLAB to analyze the mine detection device implementation performance, generating and transmitting IoT signals, and analyzing each received signal to verify the detection performance of landmines. Then we will measure the performance through the simulation of IoT-based mine detection algorithm so that we will prove the possibility of IoT-based detection landmine.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_3
• /
• pp.979-995
• /
• 2023

As wildfires are difficult to predict, real-time monitoring is crucial for a timely response. Geostationary satellite images are very useful for active fire detection because they can monitor a vast area with high temporal resolution (e.g., 2 min). Existing satellite-based active fire detection algorithms detect thermal outliers using threshold values based on the statistical analysis of brightness temperature. However, the difficulty in establishing suitable thresholds for such threshold-based methods hinders their ability to detect fires with low intensity and achieve generalized performance. In light of these challenges, machine learning has emerged as a potential-solution. Until now, relatively simple techniques such as random forest, Vanilla convolutional neural network (CNN), and U-net have been applied for active fire detection. Therefore, this study proposed an active fire detection algorithm using state-of-the-art (SOTA) deep learning techniques using data from the Advanced Himawari Imager and evaluated it over East Asia and Australia. The SOTA model was developed by applying EfficientNet and lion optimizer, and the results were compared with the model using the Vanilla CNN structure. EfficientNet outperformed CNN with F1-scores of 0.88 and 0.83 in East Asia and Australia, respectively. The performance was better after using weighted loss, equal sampling, and image augmentation techniques to fix data imbalance issues compared to before the techniques were used, resulting in F1-scores of 0.92 in East Asia and 0.84 in Australia. It is anticipated that timely responses facilitated by the SOTA deep learning-based approach for active fire detection will effectively mitigate the damage caused by wildfires.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.3 s.309
• /
• pp.30-38
• /
• 2006

In this paper, we improved to drops recognition rate according to distance using distance and depth information with 3D from stereo face images. A monocular face image has problem to drops recognition rate by uncertainty information such as distance of an object, size, moving, rotation, and depth. Also, if image information was not acquired such as rotation, illumination, and pose change for recognition, it has a very many fault. So, we wish to solve such problem. Proposed method consists of an eyes detection algorithm, analysis a pose of face, md principal component analysis (PCA). We also convert the YCbCr space from the RGB for detect with fast face in a limited region. We create multi-layered relative intensity map in face candidate region and decide whether it is face from facial geometry. It can acquire the depth information of distance, eyes, and mouth in stereo face images. Proposed method detects face according to scale, moving, and rotation by using distance and depth. We train by using PCA the detected left face and estimated direction difference. Simulation results with face recognition rate of 95.83% (100cm) in the front and 98.3% with the pose change were obtained successfully. Therefore, proposed method can be used to obtain high recognition rate with an appropriate scaling and pose change according to the distance.