Mobile Entity Detection Using Optical Flow Clustering and Vehicle Motion Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing mobile entity detection systems falsely detect stationary objects like road markings, parked vehicles, or signboards as moving entities due to their relative movement with the vehicle, leading to inaccurate detection.

Innovation Solution

A mobile entity detection apparatus that calculates unbiased variance of luminance change, extracts feature points using the FAST algorithm, performs optical flow specification and clustering, tracks overlapping clusters, and filters out background clusters based on vehicle direction and speed to accurately distinguish between mobile entities and background objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If optical flow selection based on vanishing point intersection is used to detect mobile entities, then mobile entity detection can be achieved, but false detection of stationary background objects occurs

Engineering Contradiction:
Improvemobile entity detection capabilityVSAvoiddetection accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent segments the detection process into multiple stages: initial mobile entity detection using vanishing point intersection, followed by separation of optical flows into background and foreground components, and final verification by comparing foreground optical flows with vehicle movement. This segmentation allows the system to maintain detection capability while eliminating false positives from stationary objects.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary verification mechanism using vehicle movement data as a reference. By comparing the movement characteristics of detected optical flows against known vehicle movement patterns, the system acts as a mediator to distinguish true mobile entities from false detections caused by stationary background objects.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If vanishing point-based optical flow selection is applied, then mobile entities can be detected, but stationary objects like road markings and parked vehicles are falsely detected as mobile

Engineering Contradiction:
Improvedetection coverageVSAvoidfalse detection rate
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent divides optical flows into separate background and foreground components through a dedicated separation process. Background optical flows (including stationary objects) are isolated and excluded from final detection results, while foreground optical flows representing true mobile entities are retained. This segmentation maintains comprehensive detection coverage while eliminating false detections.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and removes background optical flows from the detection process by identifying their characteristic patterns (such as movement consistent with vehicle motion only) and separating them from foreground mobile entities. This extraction process eliminates stationary objects like road markings and parked vehicles from false detection while preserving true mobile entity detection.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If comprehensive optical flow analysis is performed to improve detection accuracy, then detection precision increases, but system complexity increases

Engineering Contradiction:
Improvedetection accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the complex optical flow analysis into distinct, manageable modules: vanishing point calculation, optical flow extraction, background-foreground separation, and verification against vehicle movement. Each module performs a specific function with well-defined inputs and outputs, making the overall complex system more manageable and implementable while maintaining high detection accuracy.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3591613B1Mobile entity detector, mobile entity detection method, and computer-readable recording medium
Publication Date: 2021.12.15 NEC SOLUTION INNOVATORS LTD
  • EP3591613B1 patent drawingFigure 1
  • EP3591613B1 patent drawingFigure 2
  • EP3591613B1 patent drawingFigure 3

AI summary

A mobile entity detection apparatus 10 includes: a feature point extraction unit 11 configured to extract, from image data output by an imaging device of a vehicle, a feature point moving in an image; an optical flow specification unit 12 configured to compare the latest image data with image data that was output in the past, and specify an optical flow for each feature point; a clustering unit 13 configured to perform clustering on the optical flow based on the optical flow and the position of the optical flow in the image, and separate the optical flow into a cluster; a filtering unit 14 configured to remove a cluster that corresponds to an object forming a background from among the clusters based on the traveling direction and speed of the vehicle; and a mobile entity specification unit 15 configured to specify a mobile entity in the image using a result of clustering after cluster removal.