Estimating the number of people in crowded scenes by MID based foreground segmentation and head-shoulder detection

By Min Li, Zhaoxiang Zhang, Kaiqi Huang, Tien-Ping Tan. International Conference on Pattern Recognition, 2008

445 citations recorded by the source corpus.

This paper proposes a novel method to address the problem of estimating the number of people in surveillance scenes with people gathering and waiting. The proposed method combines a MID (mosaic image difference) based foreground segmentation algorithm and a HOG (histograms of oriented gradients) based head-shoulder detection algorithm to provide an accurate estimation of people counts in the observed area. In our framework, the MID-based foreground segmentation module provides active areas for the head-shoulder detection module to detect heads and count the number of people. Numerous experiments are conducted and convincing results demonstrate the effectiveness of our method.

Subfields: Computer Vision and Pattern Recognition, Human-Computer Interaction

View Paper: https://www.semanticscholar.org/paper/cac0bbd0b323d474253171c003b7fe100c65f05f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

Understanding Transit Scenes: A Survey on Human Behavior-Recognition Algorithms (2010, background): “Some techniques try to mitigate this issue by detecting only heads [120] or omegashaped regions formed by heads and shoulders [121].”
Scale Aggregation Network for Accurate and Efficient Crowd Counting (2018, background): “Most of the early works [13,14] estimate crowd count via pedestrian detection [15,16,17], which use body or part-based detector to locate people in the crowd image and sum them up.”
Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs (2017, methodology): “Initial research focussed on detection style [17] and segmentation framework [35].”
Learning Short-Cut Connections for Object Counting (2018, methodology): “In summary, different type of counting techniques have been applied such as counting by detection [7, 8, 9, 19, 21, 27, 39, 40], counting by clustering [30, 37], and counting by regression [3, 6, 10, 20, 28, 31, 44].”
Electronic frog eye: Counting crowd using WiFi (2014, background): “Video (or image) based recognition has been widely deployed in many public places [1].”
Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (2016, background): “In detectionbased crowd counting methods, people typically assume a crowd is composed of individual entities which can be detected by some given detectors [13, 34, 18, 10].”
Crowd Counting and Density Estimation by Trellis Encoder-Decoder Networks (2019, methodology): “Detection-based counting methods deploy a detector to traverse the image, which localizes and counts the targets along the way [5, 9, 18, 45].”
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (2018, background): “Most early work on crowd counting used detectors to detect the heads or full bodies of persons in the scene [11, 19].”
Feature Mining for Localised Crowd Counting (2012, methodology): “In counting by detection [8, 16, 24], people count is estimated through detecting instances of people.”
Bayesian Loss for Crowd Count Estimation With Point Supervision (2019, background): “rlaps and occlusions between each other; 2) perspective effects may cause large variations in human size, shape, and appearance in the image. In the past decade, a number of crowd counting algorithms [22,58,21,12,5,35,20,11,7,31] have been proposed in the literature. Recently, crowd counting methods”
Distribution Matching for Crowd Counting (2020, unknown): “For images of large crowds, this density map estimation approach has been shown to be more robust than the detection-then-counting approach [22, 19, 62, 12] because the former is less sensitive to occlusion and it does not need to commit to binarized decisions at an early stage.”
Multi-source Multi-scale Counting in Extremely Dense Crowd Images (2013, background): “Person detection for counting individuals, present in an image or video, has been employed in [10, 15].”
Iterative Crowd Counting (2018, methodology): “Li et al. [11] proposed a detection based approach where the input image was ﬁrst segmented into foreground-background regions and a HOG feature based head-shoulder detector was used to detect each person in the crowd.”
Counting Vehicles with Cameras (2018, methodology): “In the monolithic detection the classifier is trained in order to recognize the whole object we want to detect [7], while in the part-based detection we looking for specific parts of the object (such as head and shoulders for people detection) [9].”