Photoelectric Conversion Device Differential Image Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In vehicle-mounted camera systems for autonomous driving and collision prevention, low frame rates lead to slower recognition processing, which lowers recognition precision, especially under low illumination conditions where longer accumulation periods are used, increasing the risk of subject blur.
Innovation Solution
A photoelectric conversion device with a control unit and generating unit, where each pixel includes a sensor unit, counters, and memories. The control unit manages the read-out of count values from different accumulation periods within a full frame period, ensuring that a count value for a shorter accumulation period is read out before the end of a longer accumulation period, and the read-out timing differs between counters, generating a differential image based on signal differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the frame rate is increased to improve recognition speed, then recognition processing speed is improved, but the processing time becomes insufficient for completing fixed processing time requirements
Solution Approach 1:
The patent segments the full frame period into multiple accumulation periods of different lengths (first accumulation period and second accumulation period). By dividing the frame into multiple accumulation phases, the system can perform recognition processing on shorter accumulation data while maintaining the ability to use longer accumulation data when needed, thus improving recognition speed without sacrificing processing completeness.
Solution Approach 2:
The patent implements partial action by performing recognition processing on a portion of the accumulated data (first accumulation period) before the full accumulation period ends. This allows the system to start recognition processing earlier with available data, improving speed while the remaining accumulation data can be processed subsequently to ensure completeness.
2Illumination intensity
If the accumulation period is made longer to improve image brightness under low illumination, then image brightness is improved, but subject blur increases due to motion during the longer accumulation period
Solution Approach 1:
The patent dynamically adjusts the accumulation period length based on imaging conditions. By having multiple accumulation periods with different lengths (first and second accumulation periods), the system can adaptively select appropriate accumulation durations - using shorter periods to reduce blur when subjects are moving rapidly, and longer periods to improve brightness under low illumination when motion is less of a concern.
Solution Approach 2:
The patent changes the accumulation period parameter to have multiple discrete values (first accumulation period and second accumulation period with different lengths). This allows the system to optimize the balance between brightness and sharpness by selecting appropriate accumulation period lengths based on scene conditions, subject motion, and illumination levels.
3Manufacturing precision
If the frame rate is increased to reduce subject blur, then image sharpness is improved, but recognition processing cannot be completed in time due to fixed processing time requirements
Solution Approach 1:
The patent performs preliminary recognition processing on data from the first accumulation period before the second (longer) accumulation period ends. This preliminary action allows the system to start recognition processing earlier with available data, improving the timing for processing completion while maintaining image sharpness through the shorter first accumulation period that reduces subject blur.
4Productivity
If multiple accumulation periods are read out simultaneously, then data processing efficiency is improved, but read-out timing conflicts occur between different accumulation periods
Solution Approach 1:
The patent implements periodic read-out actions with different timing for different accumulation periods. By establishing distinct read-out schedules for the first and second accumulation periods, the system maintains reliable timing accuracy for each while still achieving efficient processing through the structured periodic nature of the read-out operations within the full frame period.
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
This approach enables rapid image recognition with maintained precision, reducing subject blur and allowing for obstacle recognition even at high speeds, thus preventing accidents in autonomous driving systems.
Implementation Method 1
Each pixel of the plurality of pixels includes a sensor unit that emits a pulse at a frequency corresponding to a photon reception frequency
Data Source
AI summary
A photoelectric conversion device according to one embodiment of the present disclosure has a plurality of pixels, a control unit, and a generating unit. Each pixel of the plurality of pixels includes a sensor unit that emits a pulse at a frequency corresponding to a photon reception frequency, a first counter and a second counter that count a number of the pulses, and a first memory that stores the count value for the first counter, and a second memory that stores the count value for the second counter. The control unit controls the plurality of pixels such that during a read-out operation for the count values that have been accumulated during a plurality of different accumulation periods within a full frame period from each of the first counter and the second counter, a count value for a first accumulation period is read out before an end time of a second accumulation period that is longer than the first accumulation period, and performs control such that a read-out timing for each accumulation period becomes different between the first counter and the second counter. The generating unit generates a differential image based on a difference in signals corresponding to the count values that are output from the first memory and the second memory respectively.