Event-Based Visual Tracking Without Field Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing high-speed real-time object tracking technologies face limitations in processing speed due to repetition of similar information in successive images, leading to image under-sampling or field reduction, and require complex software processing, which is not suitable for applications like micro-robotics and haptics.

Innovation Solution

A method for visual tracking of objects represented by a cloud of points, using spatio-temporal events from an asynchronous sensor, where the probability of an event belonging to the object cloud is determined, and the object's position, size, and orientation are calculated without event minimization or accumulation, allowing for robust tracking over the entire field without field reduction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If conventional camera data acquisition is used to achieve high-speed tracking, then data acquisition speed can reach tens of kilohertz, but processing speed is limited due to repetition of similar information in successive images, requiring image subsampling or field of view reduction

Engineering Contradiction:

Improvedata acquisition speedVSAvoidinformation redundancy

Core Design Contradiction:

SpeedVSLoss of information

Solution Approach 1:

The patent extracts only the relevant changes from the image data by using event-based processing. Instead of processing complete successive images, the system extracts and processes only the events (changes) that occur between frames, eliminating redundant information while preserving useful tracking data.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements dynamic processing by using an event-driven architecture where processing is triggered by actual changes in the scene rather than by fixed time intervals. This allows the system to adapt its processing rate to the actual dynamics of the tracked object, processing faster when objects move quickly and slower when they are stationary.

Inventive Principle:

Principle #15Dynamics

2Measurement precision

If complex software processing is used to improve tracking accuracy, then tracking precision can be improved, but processing speed decreases and hardware implementation becomes difficult

Engineering Contradiction:

Improvetracking accuracyVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces complex software-based image processing with a simplified hardware-friendly event processing architecture. By using event-based representation and simple probability calculations instead of complex image analysis algorithms, the system achieves both high accuracy and high processing speed that can be implemented in hardware.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the fundamental parameters of data representation from complete images to event tuples (x, y, t, polarity). This parameter transformation simplifies the processing requirements while maintaining tracking accuracy, enabling fast hardware implementation.

Inventive Principle:

Principle #35Parameter changes

3Quantity of substance

If event accumulation and minimization procedures are used to reduce data volume, then processing load is reduced, but tracking speed and temporal resolution decrease

Engineering Contradiction:

Improvedata volumeVSAvoidtracking speed

Core Design Contradiction:

Quantity of substanceVSSpeed

Solution Approach 1:

The patent processes events in a selective manner rather than accumulating all events. By using probability-based filtering and processing only relevant events for tracking, the system avoids the need for extensive event accumulation while maintaining accurate tracking, thus preserving temporal resolution and tracking speed.

Inventive Principle:

Principle #16Partial or excessive action

4Productivity

If field reduction or image subsampling is applied to improve processing speed, then processing can be performed faster, but the tracking coverage and field of view are reduced

Engineering Contradiction:

Improveprocessing speedVSAvoidfield of view

Core Design Contradiction:

ProductivityVSArea of stationary object

Solution Approach 1:

The patent applies event-based processing uniformly across the entire field of view without requiring subsampling or reduction. The event-driven architecture naturally processes only changed regions at full resolution, allowing the system to maintain complete field coverage while achieving high processing speeds through selective processing of event data.

Inventive Principle:

Principle #3Local quality

Data Source

PatentEP2989612B1Visual tracking of an object

Publication Date: 2020.11.18 CENT NAT DE LA RECH SCI (C N R S)

Figure 1~4
Figure 2~3

AI summary

The invention relates to a method for visually tracking at least one object represented by a cloud of points to which information is associated, characterised in that said method comprises the following steps: receiving (E1) data representing a set of space-time events; determining (E2) the likelihood that an event from the set belongs to the cloud of points representing the at least one object, for each event of the received set; determining (E3) whether or not an event belongs to the cloud of points depending on the likelihood determined for the event in question, for each event of the received set; updating (E4) the information relating to the cloud of points of the at least one object for each event that was determined to belong to the cloud of points in the preceding step; and calculating (E4, E5) the position, the size and the direction of the at least one object on the basis of the updated information.