Vehicle Detection Using Gradient Features and Integral Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing vehicle detection and tracking systems using motion analysis are sensitive to image noise and illumination changes, and struggle to classify detected objects accurately, especially when the camera is moving, limiting their effectiveness in real-world applications.

Innovation Solution

A method and system that utilize gradient-based features and integral images for vehicle detection, combining corner detection with boosted classifiers and temporal association to improve detection precision and tracking accuracy, even in varying lighting conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If motion analysis is used for vehicle detection, then the system can identify moving objects, but it becomes sensitive to image noise and illumination changes

Engineering Contradiction:
Improvedetection accuracyVSAvoidsensitivity to noise and illumination
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent segments the vehicle detection process into multiple independent components: corner detection identifies key geometric features, integral images compute gradient features efficiently, and boosted classifiers perform classification. This segmentation allows each component to specialize and reduces the impact of noise on the overall system.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces integral images as an intermediary data structure that pre-computes cumulative sums of pixel values. This intermediary enables efficient calculation of gradient features and histogram statistics without directly processing raw pixel data, thereby reducing sensitivity to illumination changes and noise.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If motion analysis alone is used, then the system can detect moving objects, but it cannot classify detected objects accurately

Engineering Contradiction:
Improveobject classification capabilityVSAvoidclassification accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent transforms the detection approach by changing from motion-based parameters to appearance-based parameters. Instead of relying on motion vectors, the system uses gradient-based features (edges, corners, textures) and aspect ratios that remain stable across different lighting conditions and camera motions, enabling accurate classification.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent combines multiple detection techniques into a composite system: corner detection filters identify geometric features, gradient-based integral images capture edge information, and boosted classifiers integrate these features. This composite approach leverages the strengths of each method to achieve accurate vehicle classification.

Inventive Principle:
Principle #40Composite materials

3Reliability

If a 3D polyhedral model is used for vehicle detection, then the system can track vehicles in stationary surveillance, but it fails when the camera and ego vehicle are moving

Engineering Contradiction:
Improvetracking stabilityVSAvoidperformance under camera motion
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent transitions from a static 3D polyhedral model to a dynamic detection system that adapts to camera motion. The system uses gradient-based features and integral images that are invariant to camera movement, allowing reliable vehicle detection and tracking whether the camera is stationary or moving with the ego vehicle.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2121405B1System and method for vehicle detection and tracking
Publication Date: 2016.11.30 SIEMENS CORP
  • EP2121405B1 patent drawingFigure 1
  • EP2121405B1 patent drawingFigure 2
  • EP2121405B1 patent drawingFigure 3

AI summary

A method for vehicle detection and tracking includes acquiring video data including a plurality of frames, comparing a first frame of the acquired video data against a set of one or more vehicle detectors to form vehicle hypotheses, pruning and verifying the vehicle hypotheses using a set of course-to-fine constraints to detect a vehicle, and tracking the detected vehicle within one or more subsequent frames of the acquired video data by fusing shape template matching with one or more vehicle detectors.