Edge Camera Hyperzoom Tracking for Low-Latency Traffic Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional security systems using conventional cameras often malfunction due to issues like camera failure to record video, loss of Wi-Fi connection, inability to work at night, and inefficiencies caused by large image file sizes and network latencies.
Innovation Solution
The system employs a computer vision processor in cameras to generate hyperzooms for persons or vehicles from image frames, allowing for traffic pattern tracking without network usage. This includes using a Kalman Filter to predict positions and executing mobile semantic segmentation models for attribute analytics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional security cameras transmit video signals through Wi-Fi network, then remote viewing capability is provided, but network latency and large image file sizes cause inefficiency
Solution Approach 1:
The patent segments the video processing workflow into edge computing (local object detection and tracking) and cloud computing (data storage and analysis). The camera performs local processing to generate only essential metadata (object positions, velocities, trajectories) instead of transmitting entire video frames, thereby reducing network traffic while maintaining remote access capability through a streamlined data pipeline.
Solution Approach 2:
The patent extracts only the critical information (object detection results, tracking data, metadata) from the video stream for transmission, rather than sending the complete video feed. This extraction approach eliminates redundant data transmission and focuses network bandwidth on essential security analytics.
2Loss of information
If conventional cameras record video continuously, then complete video footage is available, but large image file sizes consume excessive storage and bandwidth
Solution Approach 1:
The system extracts only meaningful objects and events from the video stream using computer vision algorithms. Instead of storing all video frames, it generates condensed metadata representing detected objects, their trajectories, and significant events, dramatically reducing data volume while preserving essential security information.
Solution Approach 2:
The patent transforms video data from raw pixel information into structured metadata parameters (object coordinates, velocities, identification tags). This parameter transformation compresses the data representation from millions of pixels per frame to concise numerical and categorical data, reducing storage requirements while maintaining analytical value.
3Loss of time
If camera processing is performed locally without network usage, then network latency is eliminated, but the camera must have sufficient computational power for object detection
Solution Approach 1:
The patent introduces an intermediary processing layer that bridges the camera and cloud infrastructure. The camera performs preliminary object detection using embedded algorithms, generates intermediate metadata, and transmits only this processed information to the cloud for further analysis. This intermediary approach balances local computational requirements with cloud capabilities.
Solution Approach 2:
The processing workload is segmented between edge devices (camera with basic detection algorithms) and cloud infrastructure (advanced analytics and storage). This segmentation allows the camera to perform essential real-time detection with minimal computational burden while leveraging cloud resources for comprehensive analysis.
4Measurement precision
If hyperzooms are stored in the camera for tracking, then detailed position data is available, but camera storage is consumed
Solution Approach 1:
The system extracts essential tracking information from hyperzoom images and stores only the extracted metadata (object positions, velocities, timestamps) in the camera's local storage. The actual hyperzoom images are either discarded or transferred to cloud storage, preserving high measurement precision while minimizing local storage consumption.
Solution Approach 2:
The patent transitions from storing two-dimensional image data to storing one-dimensional metadata arrays (coordinate sequences, velocity vectors). This dimensional reduction transforms complex image data into compact numerical representations, maintaining tracking precision while dramatically reducing storage requirements.
Data Source
AI summary
A computer vision processor of a camera generates hyperzooms for persons or vehicles from image frames captured by the camera. The hyperzooms include a first hyperzoom associated with the persons or vehicles. The computer vision processor tracks traffic patterns of the persons or vehicles while obviating network usage by the camera by predicting positions of the persons or vehicles using a Kalman Filter from the first hyperzoom. The persons or vehicles are detected in the second hyperzoom. The positions of the persons or vehicles are updated based on detecting the persons or vehicles in the second hyperzoom. The first hyperzoom is removed from the camera. Tracks of the persons or vehicles are generated based on the updated positions. The second hyperzoom is removed from the camera. Track metadata is generated from the tracks for storing in a key-value database located on a non-transitory computer-readable storage medium of the camera.