The Application And Research Of Mode Recognition And Target Tracking For Checking The Vestibula Function

Vestibular function closely related to the flight crew and the driver's spatial orientation, anti-motion sickness (car) capabilities, and dizzy spell, balance disorders and other diseases, it directly related to the safety of flying or driving.Therefore, the vestibular function testing of the flight crew and the driver get great attention around the world.Currently, the other overseas'advanced vestibular function test instruments also used pattern recognition and target-tracking technologies, however, the equipment is bulky and costliness.National primary technology of the Vestibular function test equipment currently is bio-electric eye movement recording techniques. Eye-movement signal acquisition is indirect, and measurement accuracy is low. Artifacts of bio-electric signals (such as blinking waves, muscle point wave, etc.) have a greater impact on the test results. Paste electrode is troublesome and time-consuming. It can not observe whether there is rotation nystagmus, and has lagged behind.With the continuous improvement of the computer hardware and camera cost-effective, and the development of technology related to image processing, the vestibular function test system based on target- tracking and pattern recognition will be more widely used.The topics based on pattern recognition and target-tracking theory to study the key technologies of the vestibular function test system.The main work includes the following sections in this paper:1) Achieve a high precision, robust and rapid pupil positioning method.In this method, firstly get the image histogram statistics to estimate the platform threshold, and then correct the histogram and all pixel gray value of the image. Obtain a reasonable threshold of each pixel in the image adaptively, and then carry through binary segmentation. With morphological filter remove the goal of the small structure element, and fill the holes in the segmented target; Finally carried out horizontal and vertical projection to the pupil images, and then calculate the pupil center and radius.2) Implements a real-time pupil tracking algorithm. Firstly, initialize the Kalman filter, and then predict the current state of the pupil motion with the Kalman prediction equation; Take the location predicted as the center region, and then carry out Mean shift matching in this certain region; And update the kalman prediction equation according to the current Mean shift tracking results.3) Implemented a real-time iris rotation tracking algorithm based on the feature point. Firstly, extract feature point using Harris corner detection algorithm; And then determine the size of tracking window by the Sobel edge detection, archieve corner matching using normallized correlation as the similarity criteria; At last, determine the object location of the current frame, and then update the current frame template.4) Design and implement a vestibular function tests hardware systems and virtual software systems. The computer system produce the target signal as a visual stimulus and show in the helmet display by computer instruction, and then process the real-time eye movement images captured by two analog video cameras inside the goggles; At last draw the whole eye vibration curves and output test results. The system is designed to complete a friendly interactive platform, laid the foundation for the farther analysis.