Vehicle Camera Collision Detection Using Two-Frame Neural Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Identifying the distance of objects to a vehicle requires significant memory, time, and computing resources, which can be improved.

Innovation Solution

A neural network processes two images captured at different times from a vehicle's camera to determine which objects will collide within a specified time interval by generating a binary segmentation map, enabling collision avoidance maneuvers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional methods are used to identify object distances, then measurement precision is achieved, but computing resources and time consumption increase significantly

Engineering Contradiction:

Improveobject distance measurement precisionVSAvoidcomputing resource efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the object detection process into two distinct stages: a coarse detection stage that quickly identifies potential objects in the scene, and a fine detection stage that precisely measures distances only for those identified objects. This segmentation allows the system to achieve accurate distance measurement while reducing overall computing resource consumption by avoiding intensive processing of entire scenes.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by performing full precision distance measurement only on a subset of objects that are identified as relevant in the first stage. Rather than processing all objects with high computational cost, the system selectively applies intensive processing only where necessary, thereby improving computing resource efficiency while maintaining measurement precision for critical objects.

Inventive Principle:

Principle #16Partial or excessive action

2Reliability

If traditional optical flow algorithms are used for object detection, then object motion is tracked, but processing time and computational complexity increase

Engineering Contradiction:

Improveobject detection accuracyVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent divides object detection into two phases: initial object identification using simplified methods, followed by precise distance and motion measurement only for detected objects. This segmentation reduces processing time by avoiding application of complex algorithms to entire scenes, while maintaining detection reliability through targeted precise measurement of identified objects.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary object identification before conducting detailed distance and motion analysis. By first identifying which objects are present in the scene using faster methods, the system prepares a reduced set of targets for subsequent intensive processing, thereby reducing overall processing time while ensuring detection reliability through the two-stage approach.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12485889B2Object detection and collision avoidance using a neural network

Publication Date: 2025.12.02 NVIDIA CORP

AI summary

Apparatuses, systems, and techniques to identify objects in view of a camera associated with a vehicle. In at least one embodiment, objects with which a vehicle may collide are identified, based on, for example, a difference between a size of an image of the objects detected at a first point in time and a size of an image of the objects detected at a subsequent point in time.