The Automatic Detection Algorithm And System Design Of Airborne Display Screen Symbols

In the field of aviation,in order to provide the pilot with help messages intuitively,the airborne display system in the cockpit projects the key symbols which describing the flight status and guidance messages onto the display screen in front of the pilot,after processing various messages of sensors and radar in the aircraft.Accurate symbol messages can ensure the pilot understands and controls the aircraft correctly at all times,so it is very important to verify the screen display in the cockpit to provide correct symbol messages.However,the traditional verification method depends on tester’s visual observation and experience judgment,which has problems of low accuracy and high labor costs.Therefore,it is an urgent to study the automatic detection of airborne symbols and improve the automatic testing level of airborne display system.Symbols on the airborne display screen have features of a great variety and complicated distribution.To solve this problem,this dissertation studies symbol detection algorithm combining global positioning and detail extraction,and proposes the symbol detail extraction algorithm based on image morphological component analysis,and the parallel implementation of symbol detection is carried out.The method has achieved high detection accuracy and efficiency.Combining with the current hot spots in the field of target detection,on the basis of YOLOv2 neural network,a symbol detection algorithm based on attention guided deep learning is proposed.The improved neural network is trained to learn the characteristics of various symbols on the complex picture and identify them accurately.Experimental results show that the proposed symbol detection methods can detect various types of symbols in the image accurately,and are effective for symbol detection problem in airborne display system.This dissertation proposes a symbol detection algorithm for airborne display screen which combines global positioning and detail extraction.The algorithm uses circular and radial filter to process the source image,then compares its similarity with the symbol picture in symbol database,and select eligible pixels.After positioning the target symbol,a symbol detail extraction algorithm based on morphological component analysis is designed to extract the details of each target symbol,according to the morphological structure characteristics of the target symbols.This dissertation designs a symbol detection algorithm for airborne display screen based on attention-guided neural network.Based on the structure and attention mechanism of the YOLOv2 neural network,this dissertation studies the multi-scale fusion method,designs the attention module based on the channel domain and the spatial domain,and then designs the airborne picture symbol detection algorithm of the attention-guided deep learning.The experimental results show that the symbol detection method proposed in this dissertation can accurately detect all kinds of symbols from the image,which is an effective way to solve the symbol detection problem of airborne display system.This dissertation designs and develops an automatic detection system of airborne display screen.Base on the implementation of symbol detection and symbol detail extraction algorithm of airborne display screen,modular development is carried out,and a complete automatic detection system is formed,including symbol detection module,symbol detail messages extraction module and target symbol library management module.After reasonable analysis of algorithm steps,combining with the characteristics of multi-core CPU of modern computer,the parallel algorithm is implemented based on MATLAB,which improves the speed of the detection algorithm.And the system has good stability through module test and system test.