Vehicle Object Classification Using Transparency Characteristics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated vehicles face challenges in determining whether to avoid objects based on their transparency, particularly in high-speed scenarios where abrupt braking or lane changes may be undesirable, such as with tumbleweeds, as existing systems lack effective methods to assess and respond to transparency characteristics.
Innovation Solution
An object classification system utilizing lidar and/or camera technology to determine the transparency characteristic of objects by measuring spot-distances or pixel-colors, and operating the vehicle to avoid the object if the transparency is below a threshold, ensuring safe navigation without unnecessary maneuvers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the automated vehicle performs abrupt braking or lane change to avoid all objects, then safety is improved, but unnecessary maneuvers occur when objects can be safely run over
Solution Approach 1:
The system changes the parameter of object classification by measuring transparency characteristics through lidar and camera data. By analyzing whether objects are transparent (like tumbleweeds) or opaque, the system adjusts avoidance behavior accordingly, allowing the vehicle to run over transparent objects while avoiding opaque ones, thus eliminating unnecessary maneuvers while maintaining safety
Solution Approach 2:
The transparency assessment system acts as an intermediary between object detection and avoidance decision-making. It provides additional information about object properties (transparency) that mediates the decision process, enabling the vehicle to distinguish between objects requiring avoidance and those that can be safely run over
2Strength
If the automated vehicle avoids all objects, then damage prevention is improved, but navigation efficiency deteriorates in crowded road conditions
Solution Approach 1:
The system introduces transparency as a new classification parameter to differentiate between objects requiring avoidance and those that can be safely run over. This enables the vehicle to maintain navigation efficiency by running over transparent objects like tumbleweeds while still preventing damage by avoiding opaque objects, thus resolving the contradiction between damage prevention and navigation efficiency
3Device complexity
If existing object detection systems are used without transparency assessment, then system complexity is reduced, but the ability to differentiate between run-over safe and unsafe objects is lost
Solution Approach 1:
The system achieves multi-functionality by using existing lidar and camera sensors for both standard object detection and transparency assessment. The same sensors serve dual purposes: detecting object presence and determining transparency characteristics, thus improving classification accuracy without significantly increasing system complexity
Solution Approach 2:
The transparency assessment mechanism acts as an intermediary layer that processes existing sensor data to extract additional information about object properties. This intermediary processing enhances measurement precision by providing transparency-based classification without requiring entirely new sensing systems
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Effectively allows automated vehicles to differentiate between objects that can be safely run over and those that require avoidance, optimizing navigation to prevent damage and ensure safe operation, particularly in crowded road conditions.
Implementation Method 1
The lidar determines spot-distances indicated by light-beams that were emitted by the lidar and reflected toward the lidar from an area proximate to the host-vehicle
Implementation Method 2
The camera renders an image of an area proximate to the host-vehicle. The image is based on light detected by a plurality of pixels in the camera, where each pixel detects a pixel-color of light from the area
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An object classification system (10) for an automated vehicle includes a lidar (22) and/or a camera (24), and a controller (40). The controller (40) determines a lidar-outline (42) and/or a camera-outline (68) of an object (18). Using the lidar (22), the controller (40) determines a transparency-characteristic (50) of the object (18) based on instances of spot-distances (26) from within the lidar-outline (42) of the object (18) that correspond to a backdrop-distance (48). Using the camera (24), the controller (40) determines a transparency-characteristic (50) of the object (18) based on instances of pixel-color (66) within the camera-outline (68) that correspond to a backdrop-color (72). The transparency-characteristic (50) may also be determined based on a combination of information from the lidar (22) and the camera (24). The controller (40) operates the host-vehicle (12) to avoid the object (18) when the transparency-characteristic (50) is less than a transparency-threshold (52).