雙眼影像序列對於多目標物追蹤之研究

Abstract

在軍事、保全監控等有安全顧慮的場合中，目標物的追蹤扮演著相當重要的角色。針對移動中之目標物，本論文先量測各目標物之位置，接下來，在追蹤多目標物的部分，我們利用路徑連貫函數，即時地計算出連續影像中各目標物的軌跡。當在追蹤的過程中，因遮蔽或其它因素，造成目標物的消失，本文利用α-β濾波器並且配合模糊理論，我們可以預測出各目標物下一時刻的位置，使得追蹤系統得以繼續進行。最後，利用模糊理論為工具，提出一個描述各目標物運動軌跡的方法。經由模擬的驗證，本系統具有不錯的追蹤結果。

In many occasions where safety and security are critical, such as military affairs and scene monitoring, target-tracking plays a vital role. First, in this thesis the position of the moving target is measured, and then in the subject of tracking multi-target, the trajectory of every target in an image sequence is calculated immediately by using path coherence function. Because of the occlusion or the other factors, the targets may not be detected in the tracking process. In this thesis, the next position of every target can be predicted by employing the α-β filter combined with the fuzzy theory and the tracking system will be processed. Finally, the proposed method based on the fuzzy theory can be applied to describe its motion trajectory. Moreover, the simulate results show that the tracking system has a good performance. Abstract 致謝 Contents List of Figures List of Tables 1 Introduction 1.1 Survey 1.2 Motivation 1.3 Organization of the Thesis 2 Image Processing Techniques, Greedy Exchange Algorithm and Fuzzy Systems 2.1 Image Processing Techniques 2.1.1 Image Thresholding 2.1.2 Median Filter 2.1.3 Region Growing by Pixel Aggregation 2.2 Greedy Exchange Algorithm 2.2.1 Path Coherence and Smoothness of Motion 2.2.2 Path Coherence Function 2.2.3 Gredy Exchange Algorithm 2.3 Fuzzy System 2.3.1 Fuzzifier 2.3.2 Fuzzy rule base 2.3.3 Fuzzy inference engine 2.3.4 Defuzzifer 3.System Overview of Multi-Target Tracking System 3.1 Preprocessing 3.1.1 Image Thresholding and Noise Canceling 3.1.2 Feature Extraction 3.2 3-D Coordinates Estimation 3.2.1 3-D Coordinates Computing for a Monocular CCD Camera 3.2.2 3-D Coordinates Computing for Ninocular CCD Camera 3.2.3 Using Self-Calibration Technique to Establish Binocular CCD Cameras Model 3.3 A Fuzzy Estimator 3.4 Tracking Trajectories of Feature Points 3.5 A Fuzzy Predictor 4. Simulations and Results 4.1 Simulate Environment 4.2 Simulations and Analyses 4.2.1 The Results of Estimating the Motion Trajectory 4.2.2 The Results of Tracking Multi-Target System 5.Conclusions Bibliography