Person Detection Apparatus Using Temporal Feature Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing person detection systems in monitoring cameras often misidentify static objects like posters or photographs as individuals, leading to incorrect person counts, especially in environments with high foot traffic and automatic doors, where positional changes complicate exclusion methods.
Innovation Solution
An information processing apparatus with a first determination unit to identify subjects in images and a second unit to associate regions across multiple frames, determining inclusion in the detection result based on the degree of temporal change in feature amounts, effectively distinguishing between living persons and static objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution and high frame rate images are captured to ensure suitable imaging resolution for capturing persons in high-traffic environments, then the quality of person detection is improved, but the processing load increases resulting in loss of processing efficiency
Solution Approach 1:
The patent segments the image processing task into two distinct stages: a first determination unit that performs coarse-grained processing to identify potential person regions, and a second determination unit that performs fine-grained processing to verify whether these regions correspond to actual persons or static objects. This segmentation allows high-resolution imaging without proportionally increasing overall processing load, as the second unit only processes regions identified by the first unit.
Solution Approach 2:
The patent applies partial action by having the first determination unit perform preliminary identification of person regions without complete verification, and then having the second determination unit perform verification only on these identified regions. This avoids the excessive action of applying full verification processing to all regions in high-resolution images, thereby maintaining processing efficiency while ensuring detection accuracy.
2Productivity
If determination is made on the basis of position in image to exclude persons from detection results, then processing load is reduced, but false exclusion occurs when position moves due to automatic door opening and closing
Solution Approach 1:
The patent changes the determination parameter from spatial position alone to temporal change characteristics of feature amounts. The second determination unit analyzes how feature amounts change over time to distinguish between actual persons (who exhibit temporal variation in their features as they move) and static objects (whose features remain constant despite position changes due to automatic door movement). This parameter change maintains processing efficiency while eliminating false exclusions.
3Productivity
If low load processing is applied to captured images to maintain processing efficiency, then processing speed is improved, but misdetection of static objects as persons increases
Solution Approach 1:
The patent applies preliminary action through the first determination unit, which performs preliminary identification of potential person regions before the more computationally intensive verification process. This preliminary filtering reduces the number of regions requiring detailed analysis, allowing the system to maintain low overall processing load while still achieving high detection accuracy through the two-stage approach.
Data Source
AI summary
An information processing apparatus comprising, a first determination unit configured to determine whether a subject corresponding to a person is included in an input image, and a second determination unit configured to associate a first region corresponding to a first subject determined to be included in the input image by the first determination unit with a second region determined to include the first subject in one or more images input earlier than the input image, and to determine whether the first subject is to be included in a person detection result on the basis of a degree of a temporal change of feature amounts obtained from the first region and the second region, respectively.