Autonomous Vehicle Object Detection With Field-of-View Cropping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the field of machine vision for autonomous vehicles, existing object detection methods face challenges due to high computational requirements, which necessitate downsampling of images, reducing detection accuracy and range, while faster detectors using downsampled images are not as accurate as desired.
Innovation Solution
A method that involves determining a priority field of view in an image, cropping it to high resolution, and downsampling the rest, allowing for simultaneous processing with a deep learning-based system to enhance object detection accuracy and range while reducing computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution images are processed by the detector, then detection accuracy and range are improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The image processing is segmented into two distinct paths: a first path that downsamples the entire image for fast processing, and a second path that processes only a cropped portion of the image at full resolution for accurate detection. This segmentation allows the system to balance between processing speed and detection accuracy by applying different processing levels to different parts of the image.
Solution Approach 2:
The system applies local quality enhancement by cropping a specific portion of the image that contains the object of interest and processing only that cropped region at high resolution. This allows the detector to achieve high measurement precision for the relevant area while maintaining overall processing efficiency by keeping the rest of the image at lower resolution.
2Productivity
If images are downsampled to reduce computational requirements, then processing speed is improved, but detection range and accuracy deteriorate
Solution Approach 1:
The processing pipeline is divided into two segments: a first processing path that downsamples the full image for rapid overview and object localization, and a second processing path that crops and processes the relevant region at full resolution for accurate detection. This segmentation enables the system to achieve both fast processing and high accuracy.
Solution Approach 2:
The system performs preliminary downsampling and full-image processing before cropping the image. This preliminary action allows the system to quickly identify regions of interest and prepare the cropped portion for high-resolution processing, optimizing the overall detection workflow by doing preparatory work at lower computational cost.
3Measurement precision
If the entire image is processed at full resolution, then detection accuracy is improved, but computational resources are excessively consumed
Solution Approach 1:
The system extracts only the necessary portion of the image for detailed processing by cropping the region containing the object of interest. This extraction principle allows the system to achieve high detection accuracy using only a fraction of the original image data, thereby significantly reducing computational load and energy consumption while maintaining detection precision.
Solution Approach 2:
Instead of processing the entire image at high resolution, the system applies high processing quality only to the cropped portion that contains the object of interest. This local quality approach minimizes computational resource usage by concentrating processing power where it is most needed while keeping the rest of the image at lower resolution.
4Loss of time
If downsampled images are used for detection, then processing time is reduced, but the range of detections decreases due to fewer pixels
Solution Approach 1:
The detection process is segmented into two stages: first, a downscaled image is processed quickly to identify potential objects and their locations; second, a cropped portion containing the identified objects is processed at full resolution to maintain accurate detection range. This segmentation allows the system to minimize processing time while preserving detection range for objects of interest.
Solution Approach 2:
The system performs preliminary detection on the downscaled image to quickly locate objects before cropping and processing the relevant regions at full resolution. This preliminary action enables the system to maintain short processing times while ensuring that the final detection range is preserved through subsequent high-resolution processing of the cropped portions.
Data Source
AI summary
Systems and methods for enhanced object detection for autonomous vehicles based on field of view. An example method includes obtaining an image from an image sensor of one or more image sensors positioned about a vehicle. A field of view for the image is determined, with the field of view being associated with a vanishing line. A crop portion corresponding to the field of view is generated from the image, with a remaining portion of the image being downsampled. Information associated with detected objects depicted in the image is outputted based on a convolutional neural network, with detecting objects being based on performing a forward pass through the convolutional neural network of the crop portion and the remaining portion.