Event Sensor Recognition Adjusting Reference Period by Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems using event-based sensors face challenges in maintaining recognition accuracy, particularly when the distance to the target object is long, due to insufficient information generated by the sensors, leading to decreased recognition performance.
Innovation Solution
An image recognition apparatus that includes an event data obtaining member, a metadata obtaining member, and a processing member, which adjusts the reference period for event data processing based on metadata related to the target object's position, ensuring sufficient information is gathered for accurate recognition. This apparatus uses a spiking neural network or a CNN with integration processing to adjust event data size and reference ranges, optimizing recognition accuracy and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a uniform time range for event data is used in any image region, then the processing is simple, but recognition accuracy decreases when the distance to the target object is long
Solution Approach 1:
The patent applies local quality by dividing the image into multiple regions and assigning different time ranges for event data processing to each region based on its distance from the camera. This allows each region to have optimized processing parameters tailored to its specific characteristics, thereby improving recognition accuracy for distant objects without overly complicating the overall processing system.
Solution Approach 2:
The patent implements dynamics by making the time range for event data collection adjustable rather than fixed. The system dynamically changes the time range based on the distance to the target object, allowing the processing parameters to adapt to varying conditions and maintain optimal recognition accuracy across different distances.
2Measurement precision
If the time range for event data is extended to capture more information from distant objects, then recognition accuracy improves, but processing time increases
Solution Approach 1:
The patent applies local quality by extending the time range only for regions corresponding to distant objects while maintaining shorter time ranges for nearby objects. This selective extension ensures that additional processing time is invested only where necessary (for distant objects) to improve recognition accuracy, without unnecessarily increasing processing time for the entire image.
Solution Approach 2:
The patent implements dynamics by making the time range adjustable based on object distance. The system dynamically extends the time range only when needed for distant objects, allowing the processing time to be optimized according to the specific requirements of each detection task rather than using a fixed, always-extended time range.
3Loss of information
If event data is collected for a longer period to improve recognition of distant objects, then information sufficiency increases, but system latency increases
Solution Approach 1:
The patent applies local quality by extending the data collection period only for regions corresponding to distant objects. This ensures that sufficient information is gathered for distant object recognition without unnecessarily extending the collection period for nearby objects, thereby minimizing overall system latency while maintaining information sufficiency where needed.
Solution Approach 2:
The patent implements dynamics by making the data collection period adjustable based on object distance. The system dynamically extends the collection period only when necessary for distant objects, allowing the balance between information sufficiency and system latency to be optimized for each specific detection scenario rather than using a fixed, always-extended period.
Data Source
AI summary
An image recognition apparatus includes an event data obtaining member, a metadata obtaining member, and a processing member. The event data obtaining member is configured to obtain event data that indicates changes in light amounts of pixels of a sensor. The metadata obtaining member is configured to obtain metadata related to a position of a target object. The processing member is configured to perform, using the event data, processing for recognizing the target object using an algorithm that is based on a spiking neural network. The processing member controls a period for referring to the event data in the recognition processing, based on the obtained metadata.