External Environment Recognition Apparatus for Multi-Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle detection systems face challenges in accurately detecting multiple objects within a necessary processing cycle due to increased image data and limited calculation resources, leading to inadequate detection resolution and accuracy.
Innovation Solution
An external-environment-recognizing apparatus is configured with multiple processing units, a recognition-application-storing unit, an external-information-acquiring unit, and a selecting unit that selects appropriate processing units and memory based on external information, such as vehicle behavior and environmental data, to prioritize and process image regions effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If one image sensor is shared to detect multiple objects on a time series basis, then device complexity is reduced, but detection processing cannot be completed within the necessary processing cycle
Solution Approach 1:
The patent divides the detection system into multiple dedicated processing units, each responsible for detecting specific types of objects (e.g., pedestrians, vehicles, cyclists). This segmentation allows parallel processing of different object types simultaneously, resolving the contradiction by maintaining low device complexity through specialized modules rather than a single shared sensor, while achieving the necessary detection speed through concurrent operations.
Solution Approach 2:
The patent introduces a temporal dimension by performing detection processing on time-series image data from the shared sensor. By analyzing multiple frames over time and applying motion detection algorithms, the system achieves comprehensive multi-object detection capability without requiring multiple simultaneous sensors, thus maintaining simplicity while improving detection throughput.
2Productivity
If image resolution is reduced to lower processing load, then productivity increases, but detection accuracy becomes insufficient for vehicle control
Solution Approach 1:
The patent applies local quality enhancement by selectively processing regions of interest (ROIs) at high resolution while reducing processing of background areas. The system identifies potential object locations and focuses computational resources on those specific regions, maintaining high detection accuracy for critical areas while reducing overall processing load to meet real-time requirements.
Solution Approach 2:
The patent implements a two-stage detection approach where a first detection pass identifies potential objects at lower resolution, followed by a second refined detection pass that applies higher resolution processing only to those specific regions. This partial application of high-resolution processing achieves necessary detection accuracy while keeping the overall processing load manageable within the required cycle time.
3Measurement precision
If the number of cameras and pixels is increased to improve detection accuracy, then measurement precision improves, but the processing load exceeds limited calculation resources
Solution Approach 1:
The patent makes a single image sensor perform multiple detection functions by implementing a multi-object detection algorithm that can identify different types of objects (pedestrians, vehicles, cyclists, animals) simultaneously from the same image stream. This universality eliminates the need for multiple specialized cameras while maintaining the ability to detect all object types with sufficient accuracy.
Solution Approach 2:
The patent dynamically adjusts detection parameters such as processing resolution, detection thresholds, and analysis depth based on the current scene context, object distance, and detected object types. By changing these parameters adaptively, the system maintains high detection accuracy for critical objects while reducing processing intensity for less critical areas, thereby managing the processing load within available computational resources.
Data Source
AI summary
The objective of the present invention is to obtain an external environment-recognizing apparatus with which it is possible to obtain detection results having a necessary accuracy when a plurality of target objects from a captured image are detected. This external-environment-recognizing apparatus has: a plurality of processing units for processing an image; a recognition application-storing unit for storing, by object type, a plurality of recognition applications for recognizing an object by processing using at least one processing unit among the plurality of processing units; an external-information-acquiring unit for acquiring external information that includes at least external environmental information or vehicle behavior information; and a selecting unit for selecting at least one recognition application from the recognition-application-storing unit, and for selecting, on the basis of the external information, a processing unit among the plurality of processing units for processing the selected recognition application.