Inverse Pixel Transform for Object Detection Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detectors face challenges in accurately detecting objects of varying sizes due to dependencies on distance and lens type, requiring time-consuming rescaling or increased computational resources to maintain detection accuracy.
Innovation Solution
The method involves applying an inverse pixel transform to normalize the size of objects in a target image, ensuring objects of a specific type are depicted at a consistent size regardless of lateral distance, using an image processor to prepare the image for evaluation by an object detector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the captured image is rescaled into several images having several different resolutions to enable object detection at various distances, then the object detector can correctly detect objects of different sizes, but the processing time and computational resources are significantly increased
Solution Approach 1:
The patent changes the parameter of image resolution by applying an inverse pixel transform that normalizes object sizes in the image. Instead of creating multiple images at different resolutions, the transform adjusts the pixel coordinates to make objects at various distances appear at consistent sizes, allowing a single-resolution image to be processed efficiently while maintaining detection accuracy across different object distances
2Measurement precision
If the captured image is rescaled into several images having several different resolutions to enable object detection at various distances, then the object detector can correctly detect objects of different sizes, but the computational resources are significantly increased
Solution Approach 1:
The patent transforms the image using an inverse pixel transform that modifies the spatial parameters of pixels to normalize object sizes. This single transform operation replaces the need for multiple image rescaling operations, significantly reducing computational resource requirements while maintaining the ability to detect objects at various distances with consistent accuracy
3Productivity
If the object detector is trained on one size for each object type, then the detection model is simpler and faster, but it cannot correctly detect objects of different sizes in the captured image
Solution Approach 1:
The patent applies an inverse pixel transform as a preliminary action before object detection. This transform pre-normalizes the sizes of objects in the image so that when the detector processes the image, all objects appear at the size they were trained on. This allows the detector to maintain its simple single-size training configuration while achieving accurate detection of objects at various distances
Data Source
AI summary
An image processor and a method therein to provide a target image for evaluation with an object detector. The method comprises: obtaining a source image captured by a camera and depicting an object, and applying an inverse pixel transform to each target pixel of a target image to determine one or more source pixels located at a position in the source image corresponding to a position of each target pixel in the target image. The method further comprises assigning, to each target pixel, a target pixel value determined based on one or more source pixel values of the determined one or more source pixels located at the corresponding position, thereby is a size c in the target image of the depicted object of a specific object type normalized in at least one size dimension. Thereafter, the target image is fed to an object detector for evaluation.