Vehicle Image Recognition Logic Distribution for Targeted Object Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems face challenges in cost-effectively collecting information on various objects at different locations using image recognition, as they require specialized image recognition logic for each object, which is inefficient and resource-intensive.
Innovation Solution
A server device and in-vehicle system that distributes image recognition logic to vehicles likely to encounter specific objects, allowing for cost-effective information collection without pre-installing logic for each object, using neural network models and processor-compatible image recognition logics, and dynamically distributing and deleting logic based on object recognition and vehicle location.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized image recognition logic for each object is incorporated into vehicles in advance, then recognition accuracy for specific objects is improved, but device complexity and cost increase
Solution Approach 1:
The server performs preliminary actions by selecting and distributing appropriate image recognition logic to vehicles before they need to recognize specific objects. The server identifies vehicles that will pass through regions where specific objects exist and pre-distributes the corresponding recognition logic, so that when vehicles enter these regions, they can immediately perform accurate recognition without carrying all possible recognition logics.
Solution Approach 2:
Different vehicles receive different image recognition logic based on their specific routes and the objects in their local regions. Instead of equipping all vehicles with universal recognition capabilities, the system tailors the recognition logic to each vehicle's local needs, reducing overall system complexity while maintaining recognition accuracy where needed.
2Quantity of substance
If image recognition logic is distributed to all vehicles, then information collection coverage is improved, but energy consumption and communication load increase
Solution Approach 1:
Instead of distributing image recognition logic to all vehicles, the system applies partial action by selectively distributing logic only to vehicles that will actually pass through regions where specific objects exist. This reduces unnecessary energy consumption and communication load on vehicles that don't need the recognition capability, while still achieving adequate information collection coverage through the subset of targeted vehicles.
3Reliability
If image recognition logic is retained in vehicles indefinitely, then recognition capability is maintained, but storage space and processing overhead increase
Solution Approach 1:
The image recognition logic in vehicles is made dynamic rather than static. The server can distribute, update, or delete recognition logic based on changing conditions such as vehicle route changes, object location changes, or acquisition completion. This dynamic management allows vehicles to maintain recognition capability when needed while releasing storage space when the logic is no longer required, optimizing the balance between reliability and storage efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In a server device (12), a specifying part (74) is configured to, based on information acquired from a plurality of vehicles (14), specify a distribution target vehicle to which an image recognition logic for image recognition of a predetermined object is to be distributed, among the plurality of vehicles. A distribution part (76) is configured to distribute the image recognition logic to the distribution target vehicle specified by the specifying part (74). An acquisition part (78) is configured to acquire information on the predetermined object from the distribution target vehicle to which the image recognition logic has been distributed, the information on the predetermined object being recognized by executing the image recognition logic on a captured out-of-vehicle image of the distribution target vehicle.