Hybrid Object Detection Using SSD and Pixel Difference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems, whether based on pixel difference methods or Deep Learning frameworks, face challenges in accuracy, particularly in detecting stationary objects or those moving slowly, and struggle with varying environmental conditions such as blurry images or inadequate light exposure, leading to false alarms and missed detections.
Innovation Solution
A hybrid system combining a Deep Learning framework, specifically a Single-Shot Multibox Detector (SSD) with a pixel difference framework, where the SSD generates initial predicted object locations from a single video frame and the pixel difference framework refines these using pixel differences between frames, with multiple thresholds to evaluate and filter predictions for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Deep Learning framework is used for object detection, then detection accuracy and range of conditions are improved, but computational complexity and algorithm speed are reduced
Solution Approach 1:
The patent combines Deep Learning framework (SSD) with pixel difference framework to create a hybrid system. The SSD network provides high accuracy detection while the pixel difference framework acts as a lightweight complementary system that processes motion information between frames, reducing the computational burden on the Deep Learning model while maintaining high detection accuracy across various conditions
2Productivity
If pixel difference framework is used for object detection, then algorithm speed is maintained, but detection accuracy is reduced especially for stationary or slow-moving objects
Solution Approach 1:
The system merges pixel difference framework with Deep Learning SSD network. The pixel difference framework maintains speed advantage by processing motion information efficiently, while the SSD network compensates for accuracy limitations by providing robust object detection even for stationary or slow-moving objects that pixel difference methods struggle to detect
Solution Approach 2:
The patent uses an intermediary combining mechanism that integrates predictions from both frameworks. The system combines object location predictions from SSD with motion information from pixel difference framework, using multiple thresholds to evaluate and filter predictions, thereby achieving both speed and accuracy
3Loss of time
If single frame processing is used, then computational time is reduced, but detection reliability is reduced due to false alarms and missed detections
Solution Approach 1:
The system merges single-frame processing capability of SSD with multi-frame processing of pixel difference framework. SSD processes each frame independently for fast response, while the pixel difference framework uses temporal information from multiple frames to verify detections, reducing false alarms and missed detections through cross-validation
Solution Approach 2:
The patent implements feedback mechanisms where predictions from both frameworks are evaluated against each other using multiple thresholds. The system uses feedback from pixel difference analysis to correct or confirm SSD predictions, improving reliability while maintaining acceptable processing time through efficient integration
Data Source
AI summary
An object detection system includes at least one memory storing instructions, and at least one processor that, upon executing instructions stored in the memory, controls the object detection system to perform functions including inputting a first video frame and a second video frame from a camera; generating a first set of predicted object locations using only one of the first video frame and the second video frame; generating a second set of predicted object locations based on pixel differences between the first video frame and the second video frame; and determining a final set of object locations based on the first set of predicted object locations and the second set of predicted object locations.