Fourier Fan Descriptor for Vehicle Object Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing object detection and classification systems in hazard detection and information systems face inefficiencies due to the need for multiple feature descriptors to determine 'what' and 'where' an object is, which increases computational complexity and resource requirements.

Innovation Solution

A novel feature descriptor that combines information about the feature and its location using a periodic descriptor function, allowing for a single descriptor to represent both, thereby reducing resource needs and improving detection and classification efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple feature descriptors are used to determine object characteristics and location, then detection accuracy is improved, but computational complexity and resource requirements increase

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

Solution Approach 1:

The patent combines multiple feature descriptors into a single unified descriptor that simultaneously encodes both object characteristics and spatial location information. This merging approach maintains the discriminative power of multiple descriptors while eliminating the computational overhead of processing them separately, directly resolving the contradiction between detection accuracy and computational complexity

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The single feature descriptor is designed to serve multiple functions: it simultaneously represents what the object is (class information) and where it is located (spatial information). This multi-functionality allows the system to achieve comprehensive object detection and classification using a single descriptor instead of multiple specialized descriptors, reducing resource requirements while maintaining accuracy

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If multiple feature descriptors are used to determine object characteristics and location, then detection accuracy is improved, but resource requirements increase

Engineering Contradiction:
Improvedetection accuracyVSAvoidresource requirements
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

By merging multiple feature descriptors into one unified descriptor that captures both object identity and location, the system reduces the total computational resources required for processing. This single descriptor approach maintains detection accuracy while significantly lowering energy consumption and hardware resource requirements compared to processing multiple separate descriptors

Inventive Principle:
Principle #5Merging (Combining)

3Productivity

If traditional feature descriptors are used, then object detection can be performed, but processing efficiency is reduced

Engineering Contradiction:
Improveprocessing efficiencyVSAvoiddetection time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The unified feature descriptor enables processing efficiency improvements by consolidating information that would otherwise require multiple descriptors. This merging reduces the number of computational operations needed for object detection and classification, directly improving productivity and reducing the time lost in processing

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3523749B1Object detection and classification with fourier fans
Publication Date: 2023.05.03 SMR PATENTS S A R L
  • EP3523749B1 patent drawingFigure 1
  • EP3523749B1 patent drawingFigure 2
  • EP3523749B1 patent drawingFigure 3

AI summary

An object detection and classification system includes at least one image sensor mounted on a vehicle and configured to capture an image of a portion of the environment surrounding the vehicle. The image may be stored and analyzed to detect and classify objects visible in the captured image. Keypoints are extracted from the image and evaluated according to a feature function. A new descriptor function depending on the distance in complex space between a query point and the keypoints in the image and on the feature value of the keypoints may be sampled to produce a sample value. The sample value may trigger a signal to the operator of the vehicle to respond to the object if the sample value classifies the object as satisfying a potential hazard condition.