Study On Ultrasonic Measuring Technique Of Obstacle Based On Support Vector Machine

In order to satisfy the needs of the application of obstacle measuring technique in service industry, national defence and industry, combining with the inherent favorable characteristics of ultrasonic sensor, including simple structure, mature technology, high performance-to-price ratio and good at working in the conditions of dark beam and dirt oil, a kind of ultrasonic measuring technique for obstacle based on support vector machine is developed. This technique is realized by means of classification method based on support vector machine on basis of extracting ultrasonic echo signal characters using duplicate wavelet transform and the inherent relation between signal profile and object shape. The contents of this dissertation are presented as belowChapter 1, a survey of important role of the research on obstacle ultrasonic measuring techniques is given, the review of current status and future development trend of obstacle measuring technology is summarized, the effectivity and superiority of the ultrasonic obstacle measuring are clearly indicated, the research content of this dissertation is presented and each chapter of this dissertation is arranged.Chapter 2, after statistics learning basic theory is introduced, mathematics model and theory of support vector machine in classification application are expatiated and pattern recognition technology based on support vector machine is presented.Chapter 3, the systemic scheme of ultrasonic measuring technique for obstacle is presented and the technique base on support vector machine is developed. Firstly, combining with the favorable characteristics of wavelet transform and the feature of ultrasonic echo signal, a method based on wavelet transform is used to denoise. Secondly, the technique of profile extraction based on duplicate wavelet transform is used to extract ultrasonic echo signal character owing to the trait of duplicate wavelet transform providing instantaneous phase information and the inherent relation between signal profile and object shape. Thirdly, considering conventional support vector machine aiming at two class classifier, the multi-class classifier based on support vector machine is designed especially. Finally, the performance of ultrasonic measuring technique for obstacle is evaluated and experiment results testify the validity and feasibility of this technique.Chapter 4, an embedded ultrasonic measuring system for obstacle is developed. On basis of finishing the systemic scheme, the software and hardware of the key modules are designed in details and the system integration is completed.Chapter 5, as the last chapter, the important results and conclusions of this dissertation are summarized and the prospect for the application of ultrasonic measuring technique for obstacle is put forward.