Vehicular Image Processing Combining Camera Views for Obstacle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional camera systems for vehicle obstacle detection struggle to effectively capture and detect objects at short distances due to limited field of view and installation constraints, leading to decreased detection rates and accuracy.
Innovation Solution
A vehicular image processing apparatus and system that combine images from multiple cameras with overlapping fields of view, using projective transformation to synthesize a continuous image, enabling detection of obstacles that are not entirely within a single camera's view.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a monocular camera with narrow angle of view is used to detect distant obstacles, then detection accuracy for distant objects is improved, but the feet of obstacles at short distance cannot be captured within the angle of view
Solution Approach 1:
The system divides the detection task into two segments: a monocular camera for distant obstacle detection and a downward-facing camera for short-distance obstacle detection. Each camera is optimized for its specific detection range, with the monocular camera capturing the upper body of distant objects and the downward-facing camera capturing the feet/lower body of short-distance objects. This segmentation allows both detection accuracy for distant objects and coverage for short-distance objects to be improved simultaneously.
2Area of stationary object
If a wide-angle camera is used to capture a wide area, then coverage of the ground in wide range is improved, but the feet of objects at short distance are not seen due to bonnet projection
Solution Approach 1:
The system introduces a new spatial dimension by mounting a camera on the downward-facing side of the vehicle (below the bonnet level). This downward-facing camera captures the ground area in front of the bonnet that is occluded from the perspective of forward-facing cameras. By adding this new viewing angle/dimension, the system achieves both wide coverage and accurate detection of short-distance objects' feet without being blocked by the bonnet.
3Measurement precision
If a compound-eye camera is used for distance measurement, then distance measurement capability is improved, but the feet of objects at short distance cannot be captured and template matching detection rate decreases
Solution Approach 1:
The system merges the outputs of two separate camera systems: the compound-eye camera provides distance measurement information for obstacles, while the downward-facing camera provides visual confirmation of the obstacle's feet/lower body. By combining these two data sources, the system achieves both accurate distance measurement and reliable detection of short-distance objects, overcoming the limitation where template matching fails due to missing visual information about the obstacle's feet.
4Area of stationary object
If cameras are mounted on the outer side of the vehicle for parking assistance, then wide-angle ground coverage is improved, but obstacles at short distance are difficult to detect correctly since only feet are captured
Solution Approach 1:
The system segments the obstacle detection function between two camera types: downward-facing cameras mounted on the outer side for capturing wide-angle ground views and detecting the feet of obstacles, and monocular cameras for capturing the upper body of obstacles. By segmenting the detection task and combining results from both camera systems, the system achieves both wide ground coverage and accurate obstacle detection, resolving the contradiction between coverage and detection accuracy.
Data Source
AI summary
A vehicular image processing system and a vehicular image processing apparatus are provided which combine an image from a photographing device for detection of an object forward of a vehicle and an image from a photographing device for parking assistance, thereby enabling detection, in a synthesized image, of an obstacle of which the entirety cannot be captured by each photographing device since the object is located in a region where detection is difficult in the conventional art, for example, the object is located at a short distance from the own-vehicle.