Instance-Segmented Autoencoder Processing for Vehicle Depth Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in accurately and efficiently processing depth images due to high computational demands, which can impact safety.
Innovation Solution
A method involving a binocular camera system, instance segmentation, and an autoencoder model is used to process images, including left and right images, to generate accurate depth maps by adjusting the autoencoder based on error values, enhancing image processing capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If larger and longer data calculations are performed to meet accuracy requirements and fineness of depth determinations, then measurement precision is improved, but productivity deteriorates due to increased computational time
Solution Approach 1:
The patent segments the image processing task into multiple stages: initial depth map generation, instance segmentation, error value calculation, and autoencoder model adjustment. This segmentation allows the system to process images in manageable steps, improving both precision through iterative refinement and productivity by avoiding monolithic computation
Solution Approach 2:
The patent performs preliminary actions by generating initial depth maps and instance segmentation results before final depth determination. The autoencoder model is pre-trained and adjusted based on error values calculated from preliminary processing results, enabling more efficient final computations with improved accuracy
2Measurement precision
If larger and longer data calculations are performed to meet accuracy requirements, then measurement precision is improved, but device complexity increases due to more extensive processing requirements
Solution Approach 1:
The patent introduces intermediary components including instance segmentation maps, error value calculations, and autoencoder models that mediate between raw image data and final depth determinations. These intermediaries break down the complex processing into manageable steps, reducing system complexity while maintaining precision
Solution Approach 2:
The patent implements feedback mechanisms where error values are calculated by comparing preliminary depth map results with instance segmentation data. This feedback is used to adjust the autoencoder model, creating an iterative refinement process that improves precision without requiring excessively complex single-step processing
Data Source
AI summary
A method for processing images implemented in an electronic device includes obtaining images during moving of a vehicle; obtaining instance segmentation images by segmenting the images; obtaining a predicted disparity map by reconstructing the left images based on a pre-established autoencoder; generating a first error value of the autoencoder for the images according to the left image, the predicted disparity map, and the right image, generating a second error value of the autoencoder for the instance segmentation image according to the left image of instance segmentation, the predicted disparity map, and the right image of instance segmentation; establishing an autoencoder model by adjusting the autoencoder according to the first error value and the second error value; obtaining a test image as the vehicle is moving, and obtaining a target disparity map; and obtaining a depth image corresponding to the test image by converting the target disparity map.