Drone Flight Control With Segmented Image Sensing for Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing drone flight control technologies face delays in obstacle recognition and response, particularly at high speeds, due to the sequential processing of image signals, which can lead to inadequate avoidance actions in time-sensitive situations.
Innovation Solution
A drone flight control device with an integrated image sensor and subject recognition unit that generates control signals for propeller drive based on real-time image processing, allowing for early recognition and avoidance actions by thinning pixel readout or using both image sensor and image processing unit recognition in parallel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subject recognition is performed using image processing or AI on the basis of an image signal read out from the camera, then accurate obstacle recognition is achieved, but recognition is delayed by 1 to several frames with respect to readout timing
Solution Approach 1:
The image sensor is divided into a first region and a second region. The first region performs high-speed readout for rapid obstacle detection, while the second region performs detailed image processing for accurate recognition. This segmentation allows simultaneous achievement of fast response and high accuracy without sequential processing delays.
2Reliability
If the drone changes flight direction to avoid obstacles within a safe range, then collision avoidance is achieved, but it takes time to change direction
Solution Approach 1:
The system performs preliminary obstacle detection using the first region's high-speed readout before the drone reaches critical proximity to obstacles. This early detection enables the drone to initiate direction changes in advance, ensuring reliable collision avoidance while minimizing the actual maneuvering time required.
3Measurement precision
If all pixels are read out for subject recognition, then complete image information is obtained, but recognition cannot be performed until all pixels are read out
Solution Approach 1:
The image sensor is segmented into two functional regions: the first region reads out pixels at high speed for immediate obstacle detection, while the second region processes remaining pixels for complete subject recognition. This allows the system to achieve both rapid response and comprehensive recognition accuracy simultaneously.
Solution Approach 2:
The first region performs partial pixel readout sufficient for obstacle detection purposes, enabling recognition action before complete pixel readout of all regions. This partial action approach achieves adequate recognition speed without waiting for full image data from all pixels.
Data Source
AI summary
A drone flight control device that performs flight control of a drone having a plurality of propellers, includes an image sensor including an image capturing unit that has a plurality of pixels arrayed and generates an image signal, and a first subject recognition unit that executes subject recognition processing by using the image signal, a first generation unit configured to generate a control signal for drive control of the propeller on a basis of a recognition result by the first subject recognition unit; and a drive unit configured to drive the propeller on a basis of the control signal from the first generation unit.