Side-View People Counting Using Linked Head and Body Boxes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional imaging systems struggle to accurately count people from side views due to issues with overlapping bodies, leading to inaccurate detection and low performance.
Innovation Solution
A surveillance camera system that uses bounding boxes to detect head and body regions of individuals, with a processor determining accurate counting by intersecting these boxes with preset counting lines, and linking them to avoid double counting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional image sensors use head detection from top view to count people, then detection accuracy is maintained, but applicability to side views is limited and performance deteriorates
Solution Approach 1:
The patent segments the person detection task into two distinct bounding box detections: a first bounding box for the head region and a second bounding box for the entire body region. This segmentation allows the system to handle side view images effectively by detecting both head and body separately, then determining person presence based on their spatial relationship, thereby improving adaptability to side views while maintaining detection accuracy.
2Measurement precision
If multiple bounding boxes are used to detect head and body regions, then detection coverage is improved, but device complexity increases
Solution Approach 1:
The patent merges the head detection and body detection results by establishing a linkage relationship between the first bounding box (head) and second bounding box (body). When these bounding boxes are linked, the system performs counting based on their combined information, achieving improved detection accuracy without requiring entirely separate detection systems, thus controlling device complexity.
Solution Approach 2:
The patent performs preliminary linkage determination between the first and second bounding boxes before final counting. The processor determines whether the bounding boxes originate from the same object and establishes links in advance, which simplifies the subsequent counting process and reduces computational complexity during real-time operation.
3Measurement precision
If both first and second bounding boxes are used for counting, then counting accuracy is improved, but risk of double counting increases
Solution Approach 1:
The patent performs preliminary linkage determination between the first bounding box (head) and second bounding box (body) before counting. By determining in advance whether these bounding boxes originate from the same object and establishing links, the system ensures that linked bounding boxes are counted together as a single person, thereby improving counting accuracy while preventing double counting and enhancing counting reliability.
Data Source
AI summary
A surveillance camera and a control method for the surveillance camera are disclosed. The present disclosure makes it possible to determine whether at least one of a bounding box for a head region of an objected detected from a side view image and a bounding box for an entire body region thereof intersects a preset counting line, thereby counting people passing through the counting line. In the present disclosure, one or more of a surveillance camera, an autonomous vehicle, a user terminal, and a server may be linked to an artificial intelligence module, a robot, an augmented reality (AR) device, a virtual reality (VT) device, and a 5G service.