Hardware Scanning Window for Low-Power Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer vision algorithms are resource-intensive in terms of processing power, memory usage, and data transfer bandwidth, particularly when performing feature-based tasks like face detection, which requires repetitive computations across various parameters such as location, size, scale, and rotation.
Innovation Solution
The proposed solution involves a hardware-based apparatus with a sensor array and scanning window array, coupled with peripheral circuitry and control logic, that systematically transfers and combines pixel values to efficiently store and process image data, allowing for resource-efficient computer vision computations, including the use of configurable combining circuitry for averaging pixel values and integral image computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature-based computer vision algorithms are used to perform object detection with multiple parameters (location, size, scale, rotation), then detection accuracy is improved, but processing power consumption and computation time increase significantly
Solution Approach 1:
The patent divides the image processing task into segmented operations using integral images. By pre-computing and storing cumulative sum values in a hardware array, the system segments the complex feature detection into simpler lookup and comparison operations, reducing the computational burden while maintaining detection accuracy across multiple scales and orientations
Solution Approach 2:
The patent performs preliminary computation of integral images before actual object detection. The hardware array pre-stores cumulative sum values that enable rapid feature calculation during detection, eliminating the need for repeated complex computations when evaluating multiple detection parameters
2Reliability
If feature-based algorithms are executed repeatedly with different parameters, then comprehensive object detection is achieved, but memory requirements and data transfer bandwidth increase
Solution Approach 1:
The patent creates a hardware copy of the integral image data structure that can be rapidly accessed and updated. This hardware array serves as a persistent copy of cumulative sum values, allowing the system to perform multiple detection operations with different parameters without repeatedly transferring large amounts of image data between memory and processor
3Measurement precision
If large amounts of image data are manipulated to perform computer vision algorithms, then accurate feature extraction is achieved, but data transfer bandwidth requirements increase
Solution Approach 1:
The patent extracts only the essential cumulative sum information needed for feature detection and stores it in a dedicated hardware array. By taking out and storing only the integral image data rather than manipulating complete high-resolution images repeatedly, the system reduces data transfer bandwidth requirements while maintaining feature extraction accuracy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus includes a hardware sensor array including a plurality of pixels arranged along at least a first dimension and a second dimension of the array, each of the pixels capable of generating a sensor reading. A hardware scanning window array includes a plurality of storage elements arranged along at least a first dimension and a second dimension of the hardware scanning window array, each of the storage elements capable of storing a pixel value based on one or more sensor readings. Peripheral circuitry for systematically transfers pixel values, based on sensor readings, into the hardware scanning window array, to cause different windows of pixel values to be stored in the hardware scanning window array at different times. Control logic coupled to the hardware sensor array, the hardware scanning window array, and the peripheral circuitry, provides control signals to the peripheral circuitry to control the transfer of pixel values.