Fast Human Detection for Intelligent Monitoring Using Surveillance Visible Sensors

Ko, Byoung Chul; Jeong, Mira; Nam, Jae-Yeal

doi:10.3390/s141121247

Cited by 17 publications

(9 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An RF classifier is an ensemble learning method consisting of a number of decision trees, where each tree is randomly grown with bootstrap aggregating or bagging in the training process [34]. Because an RF is based on randomizing techniques with regards to subset and feature selection while growing the trees, it is known as a classifier that is robust to overfitting, and it generates a better performance than SVM or AdaBoost-based methods [40,41].…”

Section: Facial Expression Recognition Approachmentioning

confidence: 99%

Driver’s Facial Expression Recognition in Real-Time for Safe Driving

Jeong

2018

Sensors

Self Cite

136

View full text Add to dashboard Cite

In recent years, researchers of deep neural networks (DNNs)-based facial expression recognition (FER) have reported results showing that these approaches overcome the limitations of conventional machine learning-based FER approaches. However, as DNN-based FER approaches require an excessive amount of memory and incur high processing costs, their application in various fields is very limited and depends on the hardware specifications. In this paper, we propose a fast FER algorithm for monitoring a driver’s emotions that is capable of operating in low specification devices installed in vehicles. For this purpose, a hierarchical weighted random forest (WRF) classifier that is trained based on the similarity of sample data, in order to improve its accuracy, is employed. In the first step, facial landmarks are detected from input images and geometric features are extracted, considering the spatial position between landmarks. These feature vectors are then implemented in the proposed hierarchical WRF classifier to classify facial expressions. Our method was evaluated experimentally using three databases, extended Cohn-Kanade database (CK+), MMI and the Keimyung University Facial Expression of Drivers (KMU-FED) database, and its performance was compared with that of state-of-the-art methods. The results show that our proposed method yields a performance similar to that of deep learning FER methods as 92.6% for CK+ and 76.7% for MMI, with a significantly reduced processing cost approximately 3731 times less than that of the DNN method. These results confirm that the proposed method is optimized for real-time embedded applications having limited computing resources.

show abstract

Section: Facial Expression Recognition Approachmentioning

confidence: 99%

Driver’s Facial Expression Recognition in Real-Time for Safe Driving

Jeong

2018

Sensors

Self Cite

136

View full text Add to dashboard Cite

show abstract

“…Histograms of oriented gradient (HOG) [5] is the most widely used feature descriptor for pedestrian detection. Although a dense overlapping HOG grid provides good pedestrian detection results with a lower false positive rate than traditional Haar-like descriptors, it is also produces false positives when the pedestrian is similar in color and/or pattern to the background or misses pedestrians positioned in a crowd, as well as having a heavy computation demand [6].…”

Section: Related Workmentioning

confidence: 99%

“…The role of the SOI is to determine the image-scaling level by estimating the perspective of the image and that of the ROI is to search the area of a scaled image. Ko et al [6] proposed Hough windows maps (HWMs) for determining the levels of image scaling with a divide-and-conquer algorithm to reduce the computational complexity involved in processing surveillance video sequences. Moreover, an adaptive ROI for image scaling helps improve the detection accuracy and reduce the detection time.…”

Section: Related Workmentioning

confidence: 99%

Fast Pedestrian Detection in Surveillance Video Based on Soft Target Training of Shallow Random Forest

Kim

Kwak

2019

IEEE Access

View full text Add to dashboard Cite

In recent years, deep learning algorithms have achieved top performances in object detection tasks. However, in real-time, systems having memory or computing limitations very wide and deep networks with numerous parameters constitute a major obstacle. In this paper, we propose a fast method for detecting pedestrians in surveillance systems having limited memory and processing units. Our proposed method applies a model compression technique based on a teacher-student framework to a random forest (RF) classifier instead of a wide and deep network because a compressed deep network still demands a large memory for a large amount of parameters and processing resources for multiplication. The first objective of the proposed compression method is to train a student shallow RF (S-RF), which can mimic the teacher RF's performance, by using a softened version of the teacher RF's output. Second, a deep network cannot easily detect small and closely located pedestrians in a surveillance video captured from a high perspective because of frequent convolutions and pooling processes. In this paper, adaptive image scaling and region of interest with S-RF were therefore combined to allow fast and accurate pedestrian detection in a low-specification surveillance system. In experiments, our proposed method achieved up to a 2.2 times faster speed and a 2.68 times higher compression rate than teacher RF and a better detection performance than several stateof-the-art methods on the Performance Evaluation of Tracking and Surveillance 2006, Town Centre, and Caltech benchmark datasets.

show abstract

“…Furthermore, the use of CCTV in public locations such as shopping malls, apartments, and underground parking lots has reduced the possibility of crime, including theft, assault, and/or fraud [7,8,9,10]. The use of CCTV images has expanded beyond crime prevention; for example, to ensure the safety of people on a train-station platform; to observe public-transport passengers for unexpected behaviors; and to monitor patients at hospitals [11,12,13,14]. …”

Section: Introductionmentioning

confidence: 99%

CCTV Coverage Index Based on Surveillance Resolution and Its Evaluation Using 3D Spatial Analysis

Choi

Lee

2015

Sensors

View full text Add to dashboard Cite

We propose a novel approach to evaluating how effectively a closed circuit television (CCTV) system can monitor a targeted area. With 3D models of the target area and the camera parameters of the CCTV system, the approach produces surveillance coverage index, which is newly defined in this study as a quantitative measure for surveillance performance. This index indicates the proportion of the space being monitored with a sufficient resolution to the entire space of the target area. It is determined by computing surveillance resolution at every position and orientation, which indicates how closely a specific object can be monitored with a CCTV system. We present full mathematical derivation for the resolution, which depends on the location and orientation of the object as well as the geometric model of a camera. With the proposed approach, we quantitatively evaluated the surveillance coverage of a CCTV system in an underground parking area. Our evaluation process provided various quantitative-analysis results, compelling us to examine the design of the CCTV system prior to its installation and understand the surveillance capability of an existing CCTV system.

show abstract

Fast Human Detection for Intelligent Monitoring Using Surveillance Visible Sensors

Cited by 17 publications

References 11 publications

Driver’s Facial Expression Recognition in Real-Time for Safe Driving

Driver’s Facial Expression Recognition in Real-Time for Safe Driving

Fast Pedestrian Detection in Surveillance Video Based on Soft Target Training of Shallow Random Forest

CCTV Coverage Index Based on Surveillance Resolution and Its Evaluation Using 3D Spatial Analysis

Contact Info

Product

Resources

About