Research On Characters Recognition System Of Aircraft Cockpit Overhead Panel And Indicator Lights

Many instruments and indicator lights in the cockpit of modern aircraft convey information about the status of the aircraft through characters.As a traditional technology in the field of pattern recognition,it is significant to apply it to character recognition in aircraft cockpit for the realization of artificial intelligent flight co-pilot robot.The characters of A320 aircraft cockpit overhead panel and indicator lights are taken as the research object to be recognized and a character recognition embedded system is created through the research on the system hardware platform composition and system software in this dissertation.The system hardware platform consists of Raspberry Pi 4B development board,CSI camera and 3.5-inch display.The system software includes four parts: image preprocessing,character positioning,character segmentation,and character recognition.In the image preprocessing part,the RGB color space is converted to the HSV color space to extract the amber,red,blue,and green character regions.The method of highlighting character features is selected through comparative experiments on image gray-scale and binarization methods,then the morphological operation is used to remove the noise in the character area and the skew correction is completed by the Hough transform method.In the character positioning and segmentation part,the projection method is used to complete the positioning of the character image.The binary image of the overhead panel character sequence is segmented according to the prior knowledge that the indicator characters and the overhead panel characters have fixed position information.Since the character fonts are neat,regular,and evenly distributed,the vertical projection method is used to segment the character sequence image to obtain a single character image.In the character recognition part,a support vector machine character recognition algorithm with coarse grid features combined with directional gradient histogram features is proposed for a single character.The grid search method is applied to select the optimal parameter combination of the classifier model,and the extracted character feature vectors with the optimal parameter combination is used to complete the training of the character recognition classifier model.The single characters is used to verify the effect of the classifier model,the test results show that the average accuracy of the classifier to recognize and classify a single character is over 95%.The system software is downloaded to the memory of the Raspberry Pi hardware platform,and the system is tested under the conditions of sufficient light,no shadows,and no occlusion,it can recognize the characters on the overhead panel and indicator lights.