SPAD Camera Object Tracking via Siamese Network Depth Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object tracking technologies in image processing face challenges with high data bandwidth, computation complexity, and power consumption due to reliance on RGB color space data, leading to time delays.
Innovation Solution
The method employs a single photon avalanche diode (SPAD) camera and a pre-trained siamese network to track target objects by inputting target and template images, utilizing convolutional neural networks and cross-correlation convolutional layers to determine object positions efficiently, reducing data processing and bandwidth requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If RGB color space data is used for object tracking, then tracking accuracy is improved, but data bandwidth, computation complexity, and power consumption increase
Solution Approach 1:
The patent extracts only the essential depth information from the full RGB color space data using SPAD camera technology. By capturing time-of-flight data and extracting depth maps, the system isolates the critical tracking information while discarding redundant color and texture data, thereby reducing computation complexity while maintaining tracking accuracy.
Solution Approach 2:
The patent changes the data representation parameter from RGB color space to depth map representation. This parameter transformation converts complex color information into simplified depth distance information, which reduces the dimensionality of the data and consequently lowers computation complexity and power consumption while preserving the essential spatial information needed for accurate tracking.
2Measurement precision
If RGB color space data is used for object tracking, then tracking accuracy is improved, but power consumption increases
Solution Approach 1:
The system extracts only the necessary depth information from what would otherwise be full RGB data processing. By using SPAD cameras to capture and process only depth-relevant photons, the system minimizes the computational workload and associated power consumption while maintaining the tracking accuracy required for effective object monitoring.
Solution Approach 2:
The patent transforms the data parameter from energy-intensive RGB color representation to energy-efficient depth map representation. This parameter change reduces the amount of data that needs to be processed through power-hungry neural networks, thereby significantly lowering power consumption while preserving tracking accuracy through the essential spatial depth information.
3Measurement precision
If RGB color space data is used for object tracking, then tracking accuracy is improved, but time delay increases
Solution Approach 1:
The patent extracts only the critical depth information needed for tracking from the full RGB data stream. By processing only depth maps rather than complete color images, the system reduces the computational burden and processing time, thereby minimizing time delay while maintaining tracking accuracy through the essential spatial information.
Solution Approach 2:
The patent changes the data parameter from computationally intensive RGB color space to efficient depth map representation. This parameter transformation enables faster processing through simplified neural network operations, reducing the time required for tracking calculations while preserving accuracy through the essential depth information that captures the spatial relationships needed for effective tracking.
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 significantly reduces data computation and complexity, lowering power consumption and time delays while effectively tracking target objects with reduced data size and increased processing efficiency.
Implementation Method 1
receiving at least one target image captured by a single photon avalanche diode (SPAD) camera
Data Source
AI summary
A method for tracking target object, storage medium and electronic device, which relate to the field of an image processing technology. The method includes: receiving at least one target image captured by a single photon avalanche diode (SPAD) camera before present moment; for each target image, inputting the target image and a preset template image into a pre-trained siamese network to acquire a position of a target object in the target image output by the siamese network, wherein the template image includes the target object; and determining a position of the target object in an image to be predicted based on the position of the target object in each target image, wherein the image to be predicted is an image captured by the SPAD camera at the present moment.