Omnidirectional Image Distortion for Object Detection Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing techniques, such as those described in Patent Literature 1, reduce object distortion in images to improve detection accuracy, but they cannot generate training images to accurately detect objects with increased distortion.

Innovation Solution

An image processing method that acquires an omnidirectional image, performs object detection, calculates detection accuracy, and processes the image to increase object distortion based on the detection accuracy, thereby generating a training image for accurate object detection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If image processing is performed to reduce object distortion, then detection accuracy of objects with low distortion is improved, but the system cannot generate training images for accurate detection of objects with increased distortion

Engineering Contradiction:

Improvedetection accuracyVSAvoidability to detect objects with increased distortion

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

Instead of reducing distortion as in conventional techniques, this patent applies image processing to increase distortion in training images. By inverting the traditional approach and deliberately introducing distortion through viewpoint changes, the system generates training data that enables detection models to handle distorted objects effectively.

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent changes the distortion parameter of training images by processing them with different viewpoint transformations. This parameter change allows the generation of diverse training images with varying distortion levels, enabling the detection model to adapt to different distortion conditions.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If conventional object detection processes are used on omnidirectional images, then processing speed is maintained, but detection accuracy of objects in distorted regions remains low

Engineering Contradiction:

Improveprocessing speedVSAvoiddetection accuracy in distorted regions

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary action by pre-processing omnidirectional images to create training images with increased distortion before model training. This preliminary preparation of training data with various distortion levels enables the detection model to achieve high accuracy in distorted regions without sacrificing processing speed during actual detection.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If training images with increased distortion are generated, then detection accuracy of distorted objects is improved, but image processing complexity increases

Engineering Contradiction:

Improvedetection accuracy of distorted objectsVSAvoidimage processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent uses copying by generating training images through viewpoint changes and transformations of existing omnidirectional images. Instead of creating entirely new complex processing pipelines, it copies and transforms available image data with different distortion levels, achieving improved detection accuracy while maintaining relatively simple processing.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20250118049A1Image processing method, image processing device, and non-transitory computer readable recording medium

Publication Date: 2025.04.10 PANASONIC INTELLECTUAL PROPERTY CORP OF AMERICA

AI summary

An image processing apparatus acquires an image that is made by an omnidirectional imaging and has an object associated with a truth label, executes an object detection process of detecting the object in the acquired image, calculates a detection accuracy of the object in the object detection process on the basis of the truth label, and processes the image so as to increase a distortion of the object included in the image in a case where the detection accuracy is lower than a threshold.