Automotive Sensor ROI Processing for Scenario-Based Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection and semantic segmentation methods, such as CNNs, face high processing costs and require large training datasets due to processing entire image grids, even when only specific areas are relevant for detection, leading to inefficient use of resources in varying automotive scenarios like highway and city driving.
Innovation Solution
A method that dynamically selects and adjusts the size and position of processing grids within the input grid based on prevailing vehicle scenarios, focusing processing operations on areas of interest, thereby reducing computational overhead and improving algorithm coverage by leveraging situational knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large input data grid is used to cover all potential driving scenarios, then the system can detect objects at long ranges for highway driving, but the processing overhead becomes high and computing costs increase
Solution Approach 1:
The patent divides the large input data grid into multiple smaller processing grids or regions of interest. Instead of processing the entire large grid at once, the system segments it into manageable portions that can be processed independently and in parallel, reducing the computational overhead while maintaining comprehensive coverage
Solution Approach 2:
The patent dynamically adjusts the size and position of processing grids based on the detected driving scenario. For highway driving, larger grids cover long ranges; for city driving, smaller grids focus on shorter ranges. This dynamic adaptation optimizes processing efficiency for each specific scenario while maintaining reliable object detection
2Adaptability or versatility
If the input data grid is set large enough to cover long range object detection, then highway driving scenarios are handled well, but processing overhead remains high in city driving and parking scenarios where shorter-range detection is needed
Solution Approach 1:
The system dynamically adapts the processing grid configuration based on the detected driving scenario. In highway scenarios, large grids are used for long-range detection; in city and parking scenarios, smaller grids are used for shorter-range detection. This dynamic adjustment reduces processing costs for each specific scenario while maintaining versatile scenario coverage
Solution Approach 2:
The patent applies different processing grid configurations to different spatial regions based on the driving scenario. Each region is processed with appropriate grid size and density matching the local detection requirements, optimizing energy usage while maintaining adaptability across diverse scenarios
3Adaptability or versatility
If a large training data set is provided to train the CNN, then the model can handle variety of situations, but the processing cost and data volume requirements become challenging
Solution Approach 1:
The patent segments the training process by scenario type. Instead of training on all possible scenarios uniformly, the system trains separate models or uses different configurations for different scenarios (highway, city, parking). This reduces the training data volume required for each model while maintaining overall adaptability through scenario-specific optimization
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A processing method for processing data from a sensor system, the method including the steps of receiving sensor data acquired from the sensor system including a set of data points associated with a field of view (1) of at least one sensor in an environment. Data points located within one or more areas of interest (2) are selected, the one or more areas of interest being selected based on a set of criteria. The selected data points are then processed to detect objects or perform semantic segmentation within the one or more areas of interest. The one or more areas of interest may be selected based on a scenario determination of a vehicle (10) in the environment.