• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1176-1194
• /
• 2018

The object classification using the images' contents is a big challenge in computer vision. The superpixels' information can be used to detect and classify objects in an image based on locations. In this paper, we proposed a methodology to detect and classify the image's pixels' locations using enhanced bag of words (BOW). It calculates the initial positions of each segment of an image using superpixels and then ranks it according to the region score. Further, this information is used to extract local and global features using a hybrid approach of Scale Invariant Feature Transform (SIFT) and GIST, respectively. To enhance the classification accuracy, the feature fusion technique is applied to combine local and global features vectors through weight parameter. The support vector machine classifier is a supervised algorithm is used for classification in order to analyze the proposed methodology. The Pascal Visual Object Classes Challenge 2007 (VOC2007) dataset is used in the experiment to test the results. The proposed approach gave the results in high-quality class for independent objects' locations with a mean average best overlap (MABO) of 0.833 at 1,500 locations resulting in a better detection rate. The results are compared with previous approaches and it is proved that it gave the better classification results for the non-rigid classes.

• Journal of Electrical Engineering and Technology
• /
• v.13 no.6
• /
• pp.2468-2478
• /
• 2018

We propose the Sparse Feature Convolutional Neural Network (SFCNN) to reduce the volume of convolutional neural networks (CNNs). Despite the superior classification performance of CNNs, their enormous network volume requires high computational cost and long processing time, making real-time applications such as online-training difficult. We propose an advanced network that reduces the volume of conventional CNNs by producing a region-based sparse feature map. To produce the sparse feature map, two complementary region-based value extraction methods, cluster max extraction and local value extraction, are proposed. Cluster max is selected as the main function based on experimental results. To evaluate SFCNN, we conduct an experiment with two conventional CNNs. The network trains 59 times faster and tests 81 times faster than the VGG network, with a 1.2% loss of accuracy in multi-class classification using the Caltech101 dataset. In vehicle classification using the GTI Vehicle Image Database, the network trains 88 times faster and tests 94 times faster than the conventional CNNs, with a 0.1% loss of accuracy.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.57 no.6
• /
• pp.91-97
• /
• 2015

The Unmanned Aerial Vehicles (UAVs) have several advantages over conventional RS techniques. They can acquire high-resolution images quickly and repeatedly. And with a comparatively lower flight altitude i.e. 80~400 m, they can obtain good quality images even in cloudy weather. Therefore, they are ideal for acquiring spatial data in cases of small agricultural field with mixed crop, abundant in South Korea. This paper discuss the use of low cost UAV based remote sensing for classifying crops. The study area, Gochang is produced by several crops such as red pepper, radish, Chinese cabbage, rubus coreanus, welsh onion, bean in South Korea. This study acquired images using fixed wing UAV on September 23, 2014. An object-based technique is used for classification of crops. The results showed that scale 250, shape 0.1, color 0.9, compactness 0.5 and smoothness 0.5 were the optimum parameter values in image segmentation. As a result, the kappa coefficient was 0.82 and the overall accuracy of classification was 85.0 %. The result of the present study validate our attempts for crop classification using high resolution UAV image as well as established the possibility of using such remote sensing techniques widely to resolve the difficulty of remote sensing data acquisition in agricultural sector.

• Journal of the Korean Society of Industry Convergence
• /
• v.27 no.4_2
• /
• pp.1001-1007
• /
• 2024

Recent deep learning algorithms for object detection in real-time play a crucial role in various applications such as autonomous driving, traffic monitoring, health care, and water quality monitoring. The size of small objects, in particular, significantly impacts the accuracy of detection models. However, data containing small objects can lead to underfitting issues in models. Therefore, this study developed a deep learning model capable of quickly detecting small objects to provide more accurate predictions. The RE-SOD (Residual block based Small Object Detector) developed in this research enhances the detection performance for small objects by using RGB separation preprocessing and residual blocks. The model achieved an accuracy of 1.0 in image classification and an mAP50-95 score of 0.944 in object detection. The performance of this model was validated by comparing it with real-time detection models such as YOLOv5, YOLOv7, and YOLOv8.

• Korean Journal of Remote Sensing
• /
• v.27 no.3
• /
• pp.379-388
• /
• 2011

At present, GEOBIA (Geographic Object-based Image Analysis), heir of OBIA (Object-based Image Analysis), is regarded as an important methodology by object-oriented paradigm for remote sensing, dealing with geo-objects related to image segmentation and classification in the different view point of pixel-based processing. This also helps to directly link to GIS applications. Thus, GEOBIA software is on the booming. The main theme of this study is to look into the applicability of geo-spatial processing open source to GEOBIA. However, there is no few fully featured open source for GEOBIA which needs complicated schemes and algorithms, till It was carried out to implement a preliminary system for GEOBIA running an integrated and user-oriented environment. This work was performed by using various open sources such as OTB or PostgreSQL/PostGIS. Some points are different from the widely-used proprietary GEOBIA software. In this system, geo-objects are not file-based ones, but tightly linked with GIS layers in spatial database management system. The mean shift algorithm with parameters associated with spatial similarities or homogeneities is used for image segmentation. For classification process in this work, tree-based model of hierarchical network composing parent and child nodes is implemented by attribute join in the semi-automatic mode, unlike traditional image-based classification. Of course, this integrated GEOBIA system is on the progressing stage, and further works are necessary. It is expected that this approach helps to develop and to extend new applications such as urban mapping or change detection linked to GIS data sets using GEOBIA.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.4
• /
• pp.86-97
• /
• 2011

This paper presents a novel algorithm to detect more accurate road lane with image sensor-based color classification and directional edge clustering. With treatment of road region and lane as a recognizable color object, the classification of color cues is processed by an iterative optimization of statistical parameters to each color object. These clustered color objects are taken into considerations as initial kernel information for color object detection and recognition. In order to improve the limitation of object classification using the color cues, the directional edge cures within the estimated region of interest in the lane boundary (ROI-LB) are clustered and combined. The results of color classification and directional edge clustering are optimally integrated to obtain the best detection of road lane. The characteristic of the proposed system is to obtain robust result to all real road environments because of using non-parametric approach based only on information of color and edge clustering without a particular mathematical road and lane model. The experimental results to the various real road environments and imaging conditions are presented to evaluate the effectiveness of the proposed method.

• Korean Journal of Remote Sensing
• /
• v.33 no.6_3
• /
• pp.1187-1196
• /
• 2017

Despite the continuing effort to estimate the value of function and services of ecosystem, most of the researches has used low and medium resolution satellite imagery such as MODIS or Landsat. It means that the researches to measure the ecosystem service value using VHR (Very High Resolution) satellite imagery have not been performed much, while the source of available VHR imagery is increasing. Thus, the aim of this study is to estimate and compare the result of ecosystem service value over Sejong city, S. Korea, which is one of the rapidly changed city, through the pixel-based and object-based classification analysis using VHR KOMPSAT-3 images, for more specific and precise information. In the result of the classification, forest and grassland were underestimated while agriculture and urban were overestimated in the pixel-based result compared to the object-based result. Furthermore, bare soil area was presented contrasting result that was increased in the pixel-based result, however, decreased in the object-based result. Using those results, ecosystem service values were estimated. The annual ecosystem service values in 2014 were $8.18 million USD(pixel-based) and $8.63 million USD(object-based), however, decreased to $7.80 million USD(pixel-based) and $8.62 million USD(object-based) in 2016. It is expected to use those results as a preliminary data when to make sustainable development plan and policy to improve the quality of life in the local level.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.223-228
• /
• 2009

We propose an on-line machine learning approach for object recognition, where new images are continuously added and the recognition decision is made without delay. Random forest (RF) classifier has been extensively used as a generative model for classification and regression applications. We extend this technique for the task of building incremental component-based detector. First we employ object descriptor model based on bag of covariance matrices, to represent an object region then run our on-line RF learner to select object descriptors and to learn an object classifier. Experiments of the object recognition are provided to verify the effectiveness of the proposed approach. Results demonstrate that the propose model yields in object recognition performance comparable to the benchmark standard RF, AdaBoost, and SVM classifiers.

• Journal of Korea Multimedia Society
• /
• v.19 no.12
• /
• pp.1943-1950
• /
• 2016

Recently, gender classification has attracted a great deal of attention in the field of video surveillance system. It can be useful in many applications such as detecting crimes for women and business intelligence. In this paper, we proposed a method which can detect pedestrians from CCTV video and classify the gender of the detected objects. So far, many algorithms have been proposed to classify people according the their gender. This paper presents a gender classification using convolutional neural network. The detection phase is performed by AdaBoost algorithm based on Haar-like features and LBP features. Classifier and detector is trained with data-sets generated form CCTV images. The experimental results of the proposed method is male matching rate of 89.9% and the results shows 90.7% of female videos. As results of simulations, it is shown that the proposed gender classification is better than conventional classification algorithm.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.417-422
• /
• 1998

This paper presents a method of object classification from dynamic image based on fuzzy inference algorithm which is suitable for low speed such as, conveyor, uninhabited transportation. At first, by using feature parameters of moving object, fuzzy if - then rule that can be able to adapt the wide variety of surroundings is developed. Secondly, implication function for fuzzy inference are compared with respect the proposed algorithm. Simulation results are presented to testify the performance and applicability of the proposed system.