Monocular Obstacle Detection Using Depth and Segmentation Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle computing and sensor systems are ineffective in detecting and measuring unknown objects, particularly in complex environments, leading to challenges in obstacle detection and avoidance during automated driving.
Innovation Solution
A vehicle detection system utilizing a monocular camera, a neural network model, and an optical model to detect obstacles by estimating the height and distance of unknown objects through monocular depth estimation and segmentation mapping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advanced sensor systems are used to detect unknown objects, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments the detection task into two parallel neural network components: one for generating depth maps and another for generating segmentation maps. This division allows each component to specialize in specific aspects of object detection, improving measurement precision for unknown objects while managing system complexity through functional decomposition
Solution Approach 2:
The neural network model performs multiple functions simultaneously by generating both depth maps and segmentation maps from the same input image. This multi-functionality enables the system to detect depth, size, and object categories of unknown objects using a single integrated system rather than multiple separate sensors
2Measurement precision
If advanced sensor systems are used for complex object measurements, then measurement precision is improved, but on-board computing systems become overloaded
Solution Approach 1:
The system uses a monocular camera to obtain partial depth information through monocular depth estimation, which is then enhanced by the neural network's depth map generation. This approach provides sufficient depth measurement capability for obstacle detection without requiring the full computational power of multiple active sensors like LIDAR or 3D radar
3Productivity
If parallel processing of depth maps and segmentation maps is implemented, then productivity is improved, but device complexity increases
Solution Approach 1:
The system merges the depth estimation function and segmentation function into a single neural network model that processes input images simultaneously. This unified architecture generates both depth maps and segmentation maps in parallel through shared computational resources, improving detection speed while avoiding the overhead of completely separate processing systems
Data Source
AI summary
Systems, methods, and other embodiments described herein relate to detecting obstacles from unknown objects during automated driving by a vehicle. In one embodiment, a method includes generating, from an image that includes an unknown object, a depth map from a depth estimation component and processed data from a semantic segmentation component in parallel by using a neural network model. The method also includes detecting that the unknown object is an obstacle when the unknown object satisfies criteria using an optical model according to the depth map and a segmentation map. The method also includes determining a height of the obstacle and a distance to the obstacle according to the optical model and the criteria. The method also includes adapting a vehicle plan of the automated driving according to the height.