• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.4
• /
• pp.2075-2092
• /
• 2017

Person re-identification is to match pedestrians observed from non-overlapping camera views. It has important applications in video surveillance such as person retrieval, person tracking, and activity analysis. However, it is a very challenging problem due to illumination, pose and viewpoint variations between non-overlapping camera views. In this work, we propose a viewpoint invariant method for matching pedestrian images using orientation of pedestrian. First, the proposed method divides a pedestrian image into patches and assigns angle to a patch using the orientation of the pedestrian under the assumption that a person body has the cylindrical shape. The difference between angles are then used to compute the similarity between patches. We applied the proposed method to real-time global multi-object tracking across multiple disjoint cameras with non-overlapping field of views. Re-identification algorithm makes global trajectories by connecting local trajectories obtained by different local trackers. The effectiveness of the viewpoint invariant method for person re-identification was validated on the VIPeR dataset. In addition, we demonstrated the effectiveness of the proposed approach for the inter-camera multiple object tracking on the MCT dataset with ground truth data for local tracking.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.3
• /
• pp.1121-1141
• /
• 2020

We propose a robust visual object tracking algorithm fusing a convolutional neural network tracker trained offline from a large number of video repositories and a color histogram based tracker to track objects for mixing immersive audio. Our algorithm addresses the problem of occlusion and large movements of the CNN based GOTURN generic object tracker. The key idea is the offline training of a binary classifier with the color histogram similarity values estimated via both trackers used in this method to opt appropriate tracker for target tracking and update both trackers with the predicted bounding box position of the target to continue tracking. Furthermore, a histogram similarity constraint is applied before updating the trackers to maximize the tracking accuracy. Finally, we compute the depth(z) of the target object by one of the prominent unsupervised monocular depth estimation algorithms to ensure the necessary 3D position of the tracked object to mix the immersive audio into that object. Our proposed algorithm demonstrates about 2% improved accuracy over the outperforming GOTURN algorithm in the existing VOT2014 tracking benchmark. Additionally, our tracker also works well to track multiple objects utilizing the concept of single object tracker but no demonstrations on any MOT benchmark.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.2094-2112
• /
• 2019

Focusing on the issue that the conventional Kernel Correlation Filter (KCF) algorithm has poor performance in handling scale change and obscured objects, this paper proposes an anti-occlusion and scale adaptive tracking algorithm in the basis of KCF. The average Peak-to Correlation Energy and the peak value of correlation filtering response are used as the confidence indexes to determine whether the target is obscured. In the case of non-occlusion, we modify the searching scheme of the KCF. Instead of searching for a target with a fixed sample size, we search for the target area with multiple scales and then resize it into the sample size to compare with the learnt model. The scale factor with the maximum filter response is the best target scaling and is updated as the optimal scale for the following tracking. Once occlusion is detected, the model updating and scale updating are stopped. Experiments have been conducted on the OTB benchmark video sequences for compassion with other state-of-the-art tracking methods. The results demonstrate the proposed method can effectively improve the tracking success rate and the accuracy in the cases of scale change and occlusion, and meanwhile ensure a real-time performance.

• Smart Media Journal
• /
• v.11 no.8
• /
• pp.84-92
• /
• 2022

Precision livestock monitoring promises greater management efficiency for farmers and higher welfare standards for animals. Recent studies on video-based animal activity recognition and tracking have shown promising solutions for understanding animal behavior. To achieve that, surveillance cameras are installed diagonally above the barn in a typical cattle farm setup to monitor animals constantly. Under these circumstances, tracking individuals requires addressing challenges such as occlusion and visual appearance, which are the main reasons for track breakage and increased misidentification of animals. This paper presents a framework for multi-cattle tracking in closed barns with appearance and motion models. To overcome the above challenges, we modify the DeepSORT algorithm to achieve higher tracking accuracy by three contributions. First, we reduce the weight of appearance information. Second, we use an Ensemble Kalman Filter to predict the random motion information of cattle. Third, we propose a supplementary matching algorithm that compares the absolute cattle position in the barn to reassign lost tracks. The main idea of the matching algorithm assumes that the number of cattle is fixed in the barn, so the edge of the barn is where new trajectories are most likely to emerge. Experimental results are performed on our dataset collected on two cattle farms. Our algorithm achieves 70.37%, 77.39%, and 81.74% performance on HOTA, AssA, and IDF1, representing an improvement of 1.53%, 4.17%, and 0.96%, respectively, compared to the original method.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.6
• /
• pp.107-116
• /
• 2015

In compliance with digital device growth, the proliferation of high-tech computers, the availability of high quality and inexpensive video cameras, the demands for automated video analysis is increasing, especially in field of intelligent monitor system, video compression and robot vision. That is why object tracking of computer vision comes into the spotlight. Tracking is the process of locating a moving object over time using a camera. The consideration of object's scale, rotation and shape deformation is the most important thing in robust object tracking. In this paper, we propose a robust object tracking scheme using Random Forest. Specifically, an object detection scheme based on region covariance and ZNCC(zeros mean normalized cross correlation) is adopted for estimating accurate object location. Next, the detected region will be divided into five regions for random forest-based learning. The five regions are verified by random forest. The verified regions are put into the model pool. Finally, the input model is updated for the object location correction when the region does not contain the object. The experiments shows that the proposed method produces better accurate performance with respect to object location than the existing methods.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.39 no.12
• /
• pp.1115-1123
• /
• 2011

This paper proposes an image processing algorithms for detection and tracking of aerial vehicles in short-range. Proposed algorithm detects moving objects by using image homography calculated from consecutive video frames and determines whether the detected objects are approaching aerial vehicles by the Probabilistic Multi-Hypothesis Tracking method(PMHT). This algorithm can perform better than simple color-based detection methods since it can detect moving objects under complex background such as the ground seen during low altitude flight and consider the characteristics of vehicle dynamics. Furthermore, it is effective for the flight test due to the reduction of thresholding sensitivity against external factors. The performance of proposed algorithm is verified by applying to the onboard video obtained by flight test.

• Journal of Digital Convergence
• /
• v.14 no.11
• /
• pp.143-148
• /
• 2016

There is increasing concern about security as a need for increased safety in the information industry society. However, it does not meet the needs for safety including CCTV. Therefore, in this paper, we link the processing technology using the image information to the IPS system consisting of GPS and Beacon. It designed a conventional RFID tag attached discomfort and image tracking system is limited to complement the disadvantages identifiable area. To this end, we designed a smart device and the Internet of Things convergence system and a research to ensure the accuracy and reliability of the IPS of the access control system. Finally, by leveraging intelligent video information using a PTZ camera, and set the entrant management policies it was carried out to control the situation and control. Also, by designing the integrated video tracking system, an authentication server, visualization systems were designed to establish an efficient technique for analyzing the IPS entrant behavior patterns.

• Smart Structures and Systems
• /
• v.34 no.3
• /
• pp.157-169
• /
• 2024

Accidents involving inland waterway vessels have raised concerns regarding monitoring their navigation tracks. The economical and convenient deployment of video surveillance equipment and computer vision techniques offer an effective solution for tracking vessel trajectories in narrow inland waterways. However, field applications of video surveillance systems face challenges of small object detection and the limited field of view of cameras. This paper investigates the feasibility of using multiple monocular cameras to monitor long-distance inland vessel trajectories. The one-stage CNN model, YOLOv5, is enhanced for small object detection by incorporating generalized intersection over union loss and a multi-scale fusion attention mechanism. The Bytetrack algorithm is employed to track each detected vessel, ensuring clear distinction in multiple-vessel scenarios. An inverse projection formula is derived and applied to the tracking results from monocular camera videos to estimate vessel world coordinates under potential water level changes in long-term monitoring. Experimental results demonstrate the effectiveness of the improved detection and tracking methods, with consistent trajectory matching for the same vessel across multiple cameras. Utilizing the Savitzky-Golay filter mitigates jitter in the entire final trajectory after timing-alignment merging, leading to a better fit of the dispersed trajectory points.

• Journal of Broadcast Engineering
• /
• v.25 no.5
• /
• pp.770-775
• /
• 2020

Since 360-degree ERP frame structure has location-dependent distortion, existing video surveillance algorithms cannot be applied to 360-degree video. In this paper, an activated viewport based event detection method is proposed for 360-degree video. After extracting activated viewports enclosing object candidates, objects are finally detected in the viewports. These objects are tracked in 360-degree video space for region-based event detection. The proposed method is shown to improve the recall and the false negative rate more than 30% compared to the conventional method without activated viewports.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.12
• /
• pp.1790-1796
• /
• 2021

A large amount of image data sets are required for image deep learning, and there are many differences in the method of obtaining images and constructing image data sets depending on the type of object. In this paper, we presented a method of constructing an image data set for deep learning and analyzed the performance that varies depending on the object to be tracked. We took a video by rotating the object, and then created a data set by segmenting the video using the proposed data set construction method. As a result of performance analysis, detection rate was more than 95%, and detection rate of objects with little change in shape was higher performance. It is considered that it is effective to use the data set construction method presented in this paper for a situation in which it is difficult to obtain image data and to track an object with little change in shape within a video.