Image Segmentation for Automotive Object Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Digital image processing in automotive applications is computationally expensive, and existing methods fail to efficiently filter out unnecessary information while preserving important details, leading to increased processing costs and loss of image details.

Innovation Solution

The method involves dividing an image into segments, calculating statistical parameters like expected value, standard deviation, and kurtosis for each segment, clustering, and determining background segments to filter out non-essential information, allowing for efficient processing by excluding background pixels and segments from further processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the entire image is processed for object detection and classification, then detection accuracy is improved, but processing time and computational cost increase

Engineering Contradiction:
Improveobject detection accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The image is divided into multiple segments, and statistical parameters (expected value, standard deviation, kurtosis) are calculated for each segment. Segments are then clustered into background and non-background groups based on these parameters, allowing selective processing only of non-background segments for object detection, thereby reducing processing time while maintaining detection accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Background segments are extracted and identified through clustering based on statistical parameters, then excluded from further object detection processing. This extraction of unnecessary background information reduces the data volume requiring processing while preserving important foreground objects for detection

Inventive Principle:
Principle #2Taking out (Extraction)

2Measurement precision

If the entire image is processed for object detection and classification, then detection accuracy is improved, but computational cost increases

Engineering Contradiction:
Improveobject detection accuracyVSAvoidcomputational cost
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The image is divided into multiple segments, and statistical parameters (expected value, standard deviation, kurtosis) are calculated for each segment. Segments are then clustered into background and non-background groups based on these parameters, allowing selective processing only of non-background segments for object detection, thereby reducing processing time while maintaining detection accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Background segments are extracted and identified through clustering based on statistical parameters, then excluded from further object detection processing. This extraction of unnecessary background information reduces the data volume requiring processing while preserving important foreground objects for detection

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP3979613B1Methods and systems for filtering portions of an image
Publication Date: 2024.12.04 APTIV TECHNOLOGIES AG
  • EP3979613B1 patent drawingFigure 1
  • EP3979613B1 patent drawingFigure 2
  • EP3979613B1 patent drawingFigure 3

AI summary

A computer implemented method for filtering portions of an image comprises the following steps carried out by computer hardware components: dividing the image into a plurality of segments, each segments comprising a plurality of pixels; for each of the segments, determining at least one of an expected value, a standard deviation, and a kurtosis of the plurality of pixels of the respective segment; clustering the plurality of segments into a plurality of clusters based on the at least one of the expected value, the standard deviation, and the kurtosis of the plurality of pixels of the respective segment; for each of the clusters, determining the respective cluster as belonging to a background based on a size of the respective cluster; and determining a filtered image based on the background.