Deep Learning Face Recognition for Missing Child Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing missing child search services based on face recognition face challenges due to environmental changes such as capture posture, distance, facial expressions, and ambient light, leading to recognition errors. Additionally, there is a need to quickly trace the movement path of a missing child, especially when abducted by a kidnapper using a car.
Innovation Solution
A method utilizing deep learning for face recognition that registers missing child occurrence and finding information, calculates facial feature similarity, and delivers contact information when a match is confirmed. The system also traces the movement path by analyzing CCTV images and displaying information on digital signage along the path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If typical image recognition technique is used, then the system is simple to implement, but recognition error occurs due to environmental changes
Solution Approach 1:
The patent transforms the input image through multiple parameter changes including normalization (scaling to 112x112 pixels), color space conversion (RGB to L*a*b*), and histogram equalization to enhance contrast. These parameter transformations make the recognition system robust against environmental variations while maintaining manageable complexity through systematic processing steps.
Solution Approach 2:
The patent introduces an intermediary processing stage between image capture and recognition, including histogram equalization and normalization steps. This intermediary layer acts as a mediator that standardizes varying environmental conditions before the final recognition decision, improving reliability without requiring complete system redesign.
2Reliability
If deep learning is used for face recognition, then recognition accuracy improves, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing images (normalization, histogram equalization) before the main recognition task. By preparing the data in advance with standardized formats and enhanced contrast, the deep learning model processes images more efficiently, reducing overall processing time while maintaining high accuracy.
Solution Approach 2:
The patent segments the recognition process into distinct stages: pre-processing (normalization, histogram equalization), feature extraction, and final recognition. This segmentation allows each stage to be optimized independently, with pre-processing reducing variability and the deep learning model focusing on accurate classification, thereby improving efficiency.
3Reliability
If search area is expanded to cover large distances, then the chance of finding the child increases, but information distribution becomes inefficient
Solution Approach 1:
The patent adds the dimension of prioritization to information distribution by calculating and applying priority scores based on proximity to the missing child's location. Instead of uniformly distributing information across all areas, the system ranks search locations by priority, ensuring that high-probability areas receive immediate attention while maintaining comprehensive coverage.
Solution Approach 2:
The patent changes the parameter of information distribution from uniform to prioritized based on calculated priority scores. By dynamically adjusting distribution parameters according to location priority, the system efficiently allocates attention to high-probability areas first, reducing the effective time needed to find the child while maintaining broad search coverage.
Data Source
AI summary
Disclosed herein are an apparatus and method for providing a missing child search service based on face recognition using deep learning. The method according to an embodiment may include registering missing child occurrence information including a first missing child image and information about a location from which and a time at which a child went missing when occurrence of a missing child is reported by at least one first user terminal, registering missing child finding information including a second missing child image and information about a location and a time at which a missing child was found when finding of a missing child is reported by at least one second user terminal, calculating a similarity between pieces of facial feature information extracted from the first missing child image and the second missing child image, respectively, based on deep learning, and when the similarity is equal to or greater than a predetermined threshold value, delivering contact information of the second user terminal and the missing child finding information to the first user terminal, and delivering contact information of the first user terminal and the missing child occurrence information to the second user terminal.