Research For Multi-target Tracking Methods Based On Fuzzy Information Processing In Sensor Network System

As a key technology to promote the information field into a new development stage, the related theories and applications of sensor networking have become a hot research direction. Multiple target tracking(MTT) is a key technology in distributed sensor network system(DSNS). It can obtain more complete decisions and more accurate estimations according to various data with different levels, resolutions, accuracy, dimensions, granularities, and uncertainty by using the information fusion theory. Moreover, data association is a research difficulty in MTT. Supported by the National Natural Foundations of China(No. 61271107, No. 61301074), a series of methods for engineering applications have been proposed to study the uncertainty existed in MMT based on the framework of DSNS by using the fuzzy information processing technique. These methods mainly involve: 1) track initialization and measure-to-sensor track association in the sensor node layer, 2) sensor track-to-local track association in the local node layer, 3) local track-to-system track association in the global node layer. The thesis is classified into seven chapters and their main contents are outlined as follows:In Chapter 1, the background and importance of the research on MTT in DSNS are explained; the structure model of DSNS is introduced; the status quo of the research on information fusion and MTT; the importance of the research on data association based on fuzzy information processing techniques is explained and its development is reviewed; finally, the main achievements and arrangements of the thesis are concluded.In Chapter 2, the basic techniques on MTT in sensor networks are introduced, and its related uncertainty is analyzed. The former are namely the preprocessing techniques, which mainly contain time alignment, coordinate transformation, and target localization. The latter analyses the data characteristic and the uncertainty in sensor networks.In Chapter 3, a new heterogeneous sensor-track initialization method based on the fuzzy Hough transform(FHT) is proposed for combined measurements on a platform established by a 2D radar and an IR sensor. In the proposed method, topology sequence information on different targets is introduced to classify measurements in the same time interval from heterogeneous sensors, and these measurements are used to calculate the combined measurements by the weighted average method. Then, the fuzzy Hough transform method is utilized to detect new targets. For reducing missed detections, a track confirmation rule is defined to check the detected results. Finally, the proposed method is compared with the other track initialization method based on the traditional Hough transform through a simulational experiment.In Chapter 4, a new fuzzy recursive least squares filter(FRLSF) is proposed for tracking a single maneuvering target in environments with unknown noise covariances on measurements in MTT process. Based on FRLSF, the probabilistic data association-fuzzy recursive least squares filter(PDA-FRLSF) and the generalized joint probabilistic data association-fuzzy recursive least squares filter(GJPDA-FRLSF) are proposed for tracking a single maneuvering target and multiple targets in cluttered environments, respectively.In Chapter 5, a new tracklet-to-local track association method based on the fuzzy cluster algorithm(FC) is proposed according to the characteristic on the real data transmission from sensor nodes to local nodes. In the proposed method, all tracklets in the same time interval on different targets are mapped into a set of points in parameter space by the Hough transform method, tracklet-to-local track association is realized by the maximum entropy fuzzy cluster algorithm, replacing the traditional sensor track-to-local track association methods. Finally, the validity and feasibility of the proposed method are illustrated by the experiments using the simulational data and real-test data respectively.In Chapter 6, a new tracklet-to-system track association method based on the weighted fuzzy synthetic function(w FSF) is proposed in consideration of the influence of local tracks’ reliability on track association. By introducing the local tracks’ reliability into track association, the proposed method can take the advantage of the fuzzy synthetic function as a similarity measure to make a decision on tracklet-to-system track association quickly and correctly. Finally, a simulational experiment is used to evaluate the validity of the proposed method.In Chapter 7, the summarization to the thesis is given.