Human detection and tracking with deep convolutional neural networks under the constrained of noise and occluded scenes
- PDF / 1,892,152 Bytes
- 24 Pages / 439.37 x 666.142 pts Page_size
- 86 Downloads / 176 Views
Human detection and tracking with deep convolutional neural networks under the constrained of noise and occluded scenes Ejaz Ul Haq 1
1
1
& Huang Jianjun & Kang Li & Hafeez Ul Haq
2
Received: 19 April 2020 / Revised: 31 July 2020 / Accepted: 11 August 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Human detection and tracking is a key aspect in surveillance system due to its importance in timely identification of person, recognition of human activity and scene analysis. Convolutional neural networks have been widely used approach in detection and tracking related tasks. In this paper, a robust framework is presented for the human detection and tracking in noisy and occluded environments with the aid of data augmentation techniques. In addition, a softmax layer and integrated loss function is used to improve the detection and classification performance of the proposed model. The primary focus is to perform the human detection task in unconstrained environments. The implemented system outperforms the state-of-the-arts methods which can be validated from the experimental results. Keywords Human Detection . Deep learning . Data augmentation techniques
1 Introduction Human detection and tracking is well known problem that has been studied frequently in the computer vision literature. Automatic people detection is a vital task that has many important practical applications in video surveillance system, transportation control, security system and managing crowded scenes [47]. Due to its wide applicability, it can be divided into two aspects: human verification and human identification. The anterior focus on determining whether a human exist in a set of different objects in an image and video or not, while the human identification aim to recognizes and locate the person amongst a set of various objects * Huang Jianjun [email protected]
1
ATR Key Laboratory, School of Electronics and Information Engineering, Shenzhen University, Shenzhen, China
2
Fujian Normal University, Fuzhou, China
Multimedia Tools and Applications
[51]. Despite of enough research done in the field of human identification still it is considered a challenging task in many real time applications. Human poses are a non-rigid appearance which differentiates divergent groups of people from each other in terms of size, shape and body structure. In addition, noise, occlusions and surrounding environment greatly affect the human detection and tracking performance in a real time scenario [26]. In the recent decades, deep learning based methods are extensively used for the human identification in computer vision. Convolutional neural network has emerged as a most prevalent method for figuring out real time problems. Different vigorous and discriminative configurations of convolutional neural networks have applied in solving image processing, passenger flow calculation, crowd counting and object detection tasks [28]. In fact, neural network is a sub category of artificial intelligence having ability to learn eff
Data Loading...
