Object Recognition Target Selection Under CPU Processing Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems require extensive processing time to sort and select a set number of mobile objects based on predetermined priorities, especially when dealing with a large number or variety of objects, exceeding the processing capabilities of resources like CPUs.
Innovation Solution
An object recognition device and method that utilizes an object detection unit, a processing target object storage unit, and an object processing unit to associate evaluation values with detected objects, allowing for simpler information processing by selecting and storing objects with the lowest evaluation value, thereby reducing the need for priority sorting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detected mobile objects are sorted in accordance with predetermined priorities and then a set number of mobile objects are selected, then the most relevant objects can be identified for safety-related driving assistance control and lane change assistance control, but the information processing time becomes excessively long
Solution Approach 1:
The patent divides the object selection process into two independent stages: first selecting objects for safety-related driving assistance control based on collision risk, and second selecting objects for lane change assistance control based on lateral position. This segmentation eliminates the need for comprehensive sorting of all detected objects, significantly reducing processing time while maintaining selection accuracy for each control type.
Solution Approach 2:
The patent extracts and processes only the necessary objects for each specific control function rather than sorting all detected objects. For safety control, only objects with collision risk are extracted and processed; for lane change control, only objects in the lateral direction are extracted. This extraction approach reduces the dataset size and processing time while preserving the quality of selected objects.
2Adaptability or versatility
If a large number or variety of objects are recognized, then comprehensive situation awareness is achieved, but the processing time increases beyond CPU capabilities
Solution Approach 1:
The patent segments the comprehensive object recognition task into function-specific subtasks: safety-related object recognition and lane change object recognition. Each segment processes only the objects relevant to its function, maintaining comprehensive situation awareness while dividing the computational burden to match CPU processing capabilities.
Solution Approach 2:
The patent applies partial action by processing only the necessary subset of objects for each control function rather than processing all detected objects. For safety control, processing is limited to objects with collision risk; for lane change control, processing is limited to laterally positioned objects. This partial processing approach maintains productivity within CPU limits while achieving sufficient comprehensiveness for each function.
Data Source
AI summary
An object recognition device of the present invention includes: an object detection unit which detects, as an object, a thing present around a sensor platform; a processing target object storage unit which stores the object as a processing target object, with a number there of being an upper limit value or smaller; and an object processing unit which, in a case where the processing target objects whose number is the upper limit value are stored, selects an object having a lowest evaluation value among the stored processing target objects whose number is the upper limit value, determines one object of the selected processing target object and an object newly detected by the object detection unit, and stores the determined object as the processing target object in the processing target object storage unit.