Moving Object Detection Using Eccentricity Masks and Sparse CNNs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current traffic infrastructure systems face challenges in efficiently and quickly determining the locations of moving objects in images, which hampers the ability to operate vehicles autonomously or semi-autonomously, as existing methods require significant computing resources and time.
Innovation Solution
The implementation of eccentricity analysis to segment images and convert them into a compressed dense matrix, combined with a sparse convolutional neural network, reduces the computational load by focusing only on moving objects, allowing for faster detection and object classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional convolutional neural networks are used to detect objects in images, then object detection accuracy is maintained, but computing resources and processing time increase significantly
Solution Approach 1:
The patent segments the image processing task by first identifying moving objects through eccentricity analysis, then applying convolutional neural networks only to those specific regions rather than the entire image. This segmentation approach maintains detection accuracy for moving objects while dramatically reducing the computational burden by excluding static background areas from intensive processing.
Solution Approach 2:
The patent extracts and isolates moving objects from the full image using eccentricity maps and masking techniques. By taking out only the relevant moving object regions and processing those separately with CNNs, the system achieves accurate detection of moving objects without the computational overhead of processing the complete image, thus resolving the contradiction between accuracy and processing speed.
2Reliability
If full image processing is applied to detect all objects, then comprehensive object detection is achieved, but computing resources and time consumption increase
Solution Approach 1:
The patent introduces dynamic adaptivity by using eccentricity analysis to identify moving objects and adjusting the processing strategy accordingly. Instead of applying static full-image processing to all frames, the system dynamically determines which regions require intensive CNN processing based on motion detection, thereby reducing processing time while maintaining reliable detection of moving objects.
Solution Approach 2:
The patent applies partial processing by using eccentricity analysis as a preliminary filtering stage that identifies only the necessary regions for full CNN processing. This partial action approach processes only moving object regions with intensive algorithms while using lighter eccentricity-based detection for the remainder, achieving reliable moving object detection without the time cost of complete image processing.
3Productivity
If eccentricity analysis and sparse convolutional neural network are used, then processing time and computing resources decrease, but system complexity increases
Solution Approach 1:
The patent merges multiple processing stages into an integrated pipeline: eccentricity analysis, mask generation, region extraction, and sparse CNN processing are combined into a unified workflow. This merging consolidates the complexity into a coordinated sequence of operations that work together efficiently, achieving high processing productivity while managing system complexity through integrated design rather than separate independent components.
Data Source
AI summary
A segmentation mask can be determined that includes at least one moving object in a plurality of images based on determining eccentricity for each pixel location in the plurality of images. A first image included in the plurality of images can be segmented by applying the segmentation mask to the image. The segmented first image can be transformed to a compressed dense matrix which includes pixel values for non-zero portions of the segmented first image. The compressed dense matrix can be input to a sparse convolutional neural network trained to detect objects. A detected object corresponding to the at least one moving object included in the first image can be output from the sparse convolutional neural network.