Design And Implementation Of Golden Snub- Nosed Monkey Tracking Algorithm And Software In Natural Scene Video

Golden Snub-nosed Monkey is the first-grade state protection animal.It is necessary to protecting golden snub-nosed monkeys without any doubt.For the protection of golden snub-nosed monkeys,effective monitoring means are important measures.Digital image technology has been widely applied due to its advantages of non-invasive,non-invasive and non mandatory in the monitoring of golden snub-nosed monkey.We can get rich image data by using this technology to monitor it.Obviously,the traditional artificial processing and analysis methods need a lot of manpower and long time to analyze amount of image data,which causes the data cannot be processed in time and seriously reducing the practical application value.Therefore,the appearance of the automation monitoring is essential to address this problem.Nevertheless,how to detect and track wild golden snub-nosed monkeys in the natural environment is the first priority.However,through the analysis of the golden snub-nosed monkey video,there are still a series of problems in the natural scene of the golden snub-nosed monkey tracking.On the one hand,golden snub-nosed monkeys survive in the wild where the distribution area is wideand different areas have various enviroment.On the other hand,the golden snub-nosed monkey has quick and swift athletic talent,so the non rigid transformation caused by the golden snubnosed monkey movement is another outstanding problem in the golden snub-nosed monkey tracking process.To solve the above problems,the main work of this thesis is as follows:(1)Aiming at the problem of tracking failure caused by TLD's vulnerability to complex environmental conditions,we propose a single object tracking algorithm based on deep learning object detector PVANET(Performance Vs Accuracy Net)and TLD(TrackingLearning-Detection)for golden snub-nosed monkey,referred to as PTLD.For the first frame of the image sequence,the PVANET detector is used to determine the position of the golden snub-nosed monkey and the TLD detector is trained.If the current image frame is not the first frame and the object is not lost in the previous frame,the TLD algorithm is used to track the golden snub-nosed monkey.If the current frame is not the first frame,but the object is lost in the previous frame,the object position is re-determined using the PVANET detector,the TLD's detector is updated,and the TLD continues to be used for tracking.The experimental results show that PTLD combines the advantages of PVANET and TLD,with high speed and accuracy,and can perform single-object tracking of golden snub-nosed monkey in natural scene video.(2)In order to track all the golden snub-nosed monkeys in the field of vision,a golden snubnosed monkey multi-object tracking algorithm based on PVANET and KCF(Kernel Correlation Filter)is proposed in this thesis.The algorithm first obtains the accurate object detection result through the object detector PVANET for multi-object detection.Then,create a KCF tracker for each candidate box,in which case each object is tracked and their state is updated frame by layer.The algorithm solves the problem of scale change,target crossover,occlusion,and target loss by performing data association and target state update between the KCF tracker and the PVANET detector.The experimental results show that the golden snubnosed monkey multi-object tracking algorithm based on PVANET and KCF can perform multi-object tracking of the golden snub-nosed monkey in the natural scene video.(3)This thesis designs and implements the golden snub-nosed monkey tracking software based on Qt design,which is combined with the research results of single object algorithm and mutigarget tacking algorithm in nature scene.The software was deployed on PC platform and Jetson TX1.The software has developed a file operation module,golden snubnosed monkey single object tracking module,a golden snub-nosed monkey multi-object tracking module,and a result display module.The software achieves the tracking of golden snub-nosed monkey in natural scene video.Finally,we verified the stability of the software by actual scenario testing.