CNN Object and Lane Detection System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional convolutional neural networks (CNNs) are unable to simultaneously perform object detection, including classification and localization, and lane detection, requiring multiple parallel computational steps and increased computational time and memory.
Innovation Solution
A CNN system that includes convolution and pooling layers, a fully connected layer, and a non-maximum suppression layer, trained on annotated images for both object and lane marking detection, allowing for simultaneous object and lane detection with confidence values and localization information, and a lane line module to fit a second-order polynomial through center points of lane bounding boxes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CNNs perform object detection and lane detection using multiple parallel computational steps, then detection accuracy is maintained, but computational time increases and memory requirements increase
Solution Approach 1:
The patent combines object detection and lane detection into a single CNN model with shared convolutional layers. The network processes both detection tasks simultaneously through unified feature extraction layers, eliminating the need for separate parallel computational steps while maintaining detection accuracy through multi-task learning mechanisms.
Solution Approach 2:
The CNN model is designed as a universal system that performs multiple detection functions (object detection and lane detection) within a single architecture. The shared layers extract general features useful for both tasks, while task-specific branches handle individual detection requirements, reducing overall computational time and memory usage.
2Reliability
If conventional CNNs use multiple parallel computational steps for object detection and lane detection, then task performance is maintained, but memory requirements increase
Solution Approach 1:
The patent merges object detection and lane detection into a single computational framework with shared convolutional layers. This consolidation reduces memory requirements by eliminating redundant feature extraction processes while maintaining task performance through a unified model that learns shared representations for both detection tasks.
Solution Approach 2:
The network architecture segments the detection tasks into shared feature extraction layers and task-specific prediction branches. This segmentation allows efficient memory utilization by computing common features once and reusing them for both object and lane detection, rather than maintaining separate computational pipelines.
3Adaptability or versatility
If multiple parallel computational steps are used for object detection and lane detection, then detection capabilities are comprehensive, but device complexity increases
Solution Approach 1:
The patent implements a universal CNN architecture that handles both object detection and lane detection within a single system. This multi-functional design reduces device complexity by consolidating multiple detection capabilities into one integrated model, while maintaining comprehensive detection capabilities through shared and task-specific layers.
Solution Approach 2:
The system merges object detection and lane detection into a unified computational pipeline. By combining these detection capabilities in a single network architecture with shared layers, the patent reduces system complexity while preserving comprehensive detection functionality through multi-task learning.
Data Source
AI summary
A system and method for predicting object detection and lane detection for a motor vehicle includes a convolution neural network (CNN) that receives an input image and a lane line module. The CNN includes a set of convolution and pooling layers (CPL's) trained to detect objects and lane markings from the input image, the objects categorized into object classes and the lane markings categorized into lane marking classes to generate a number of feature maps, a fully connected layer that receives the feature maps, the fully connected layer generating multiple object bounding box predictions for each of the object classes and multiple lane bounding box predictions for each of the lane marking classes from the feature maps, and a non-maximum suppression layer generating a final object bounding box prediction for each of the object classes and generating multiple final lane bounding box predictions for each of the lane marking classes.