Predicted Object Region Detection for Low-Load Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face challenges in achieving high throughput for object detection without increasing processing load, leading to issues such as increased device usage, cost, and power consumption.
Innovation Solution
An image processing device that predicts object positions using past input images, performs partial region object detection, and determines target regions for focused detection, utilizing multiple detection units to manage processing loads effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If object detection processing is performed on the entire input image, then detection accuracy is maintained, but processing load increases
Solution Approach 1:
The patent divides the input image into multiple partial regions and performs object detection only on selected regions rather than the entire image. The object detection target region determination unit identifies specific regions based on predicted object positions from past images, and the first object detection unit processes only those regions. This segmentation approach reduces processing load while maintaining detection accuracy for expected objects.
Solution Approach 2:
The patent performs preliminary object position prediction using past input images before processing the current image. The object position prediction unit predicts where objects are likely to appear in the new image based on their positions in previous frames. This preliminary action enables the system to focus computational resources on predicted regions, reducing overall processing load while maintaining throughput.
2Productivity
If frame rate is increased to prevent detection omission, then throughput improves, but processing load increases
Solution Approach 1:
By segmenting the image into partial regions and only processing predicted target regions, the system can maintain high frame rates without proportionally increasing processing load. The object detection target region determination unit selects specific regions based on prediction results, significantly reducing the computational burden per frame while maintaining the ability to detect objects promptly.
Solution Approach 2:
The preliminary prediction of object positions from past images allows the system to prepare detection regions in advance before processing each new frame. This enables faster processing by avoiding the need to scan the entire image at high frame rates, thus maintaining throughput while reducing per-frame processing load.
3Productivity
If multiple detection units are used to increase throughput, then processing capacity improves, but device cost increases
Solution Approach 1:
The patent segments the detection process into functional units: object position prediction unit, object detection target region determination unit, second object detection unit, and first object detection unit. These units work cooperatively to process images efficiently, achieving high throughput without requiring multiple separate detection devices. The segmentation of functions enables better resource utilization.
Solution Approach 2:
The system uses a single detection device that performs multiple functions through different units: prediction, target region determination, and object detection. The first and second object detection units can operate sequentially or in coordination, with the second unit performing initial detection and the first unit performing focused detection on predicted regions. This multi-functionality achieves throughput improvement without proportionally increasing device count or cost.
Data Source
AI summary
An image processing device includes an object position prediction unit configured to predict a position of an object in a new input image based on a position of the object detected in a past input image, a second object detection unit configured to perform object detection within a partial region in the new input image, an object detection target region determination unit configured to determine an object detection target region to be an object detection target in the new input image based on a prediction result by the object position prediction unit and an object detection result by the second object detection unit, and a first object detection unit configured to perform object detection for the object detection target region determined by the object detection target region determination unit.