Stereoscopic Object Detection via Epipolar Constraints

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current electronic devices face computational intensity in detecting and segmenting objects in stereoscopic images, which hinders efficient disparity estimation and object segmentation processes.

Innovation Solution

A method that involves receiving and processing stereoscopic images by detecting object points, determining search ranges, calculating disparity values, and segmenting objects based on cost functions to assign object and non-object labels, thereby reducing computational complexity through epipolar constraints and graph cut algorithms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If object detection and segmentation is performed on stereoscopic images using conventional methods, then object detection capability is achieved, but computational complexity becomes intensive

Engineering Contradiction:
Improveobject detection accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the image processing task into distinct phases: epipolar line generation, candidate point selection along epipolar lines, disparity calculation, and cost function evaluation. This segmentation allows each phase to be optimized independently, reducing overall computational complexity while maintaining detection accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by focusing computational resources on relevant regions only. By constraining the search to epipolar lines and selecting candidate points based on local image features, the method reduces the search space from the entire image to specific localized regions, thereby reducing computational complexity without sacrificing detection precision.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If comprehensive object segmentation is performed using disparity values, then segmentation accuracy is improved, but processing time increases

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary actions by pre-calculating epipolar lines and pre-selecting candidate points before the actual segmentation process. This preliminary processing organizes the data structure in advance, allowing the segmentation algorithm to operate more efficiently on pre-processed information, thus reducing processing time while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies partial action by evaluating cost functions only for selected candidate disparity values rather than all possible disparities. By identifying a limited set of candidate points and evaluating only relevant disparity values, the method achieves sufficient segmentation accuracy with reduced processing time compared to exhaustive search methods.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3036901B1Method, apparatus and computer program product for object detection and segmentation
Publication Date: 2019.01.30 NOKIA TECHNOLOGIES OY
  • EP3036901B1 patent drawingFigure 1
  • EP3036901B1 patent drawingFigure 2
  • EP3036901B1 patent drawingFigure 3

AI summary

In an example embodiment, a method, apparatus and computer program product are provided. The method includes facilitating receipt of a first image and a second image of a scene comprising one or more objects. The method includes detecting the objects in the first image by detecting object point of the objects in the first image. The method includes detecting the object points of the objects in the second image based on detection of the object points of the objects in the first image. Detection of an object point in the second image that corresponds to an object point of the first image comprises searching for the object point on an epipolar line in the second image corresponding to the object point of the first image. The method includes determining disparity values between the objects points in the first image and the object points in the second image.