Presentation is loading. Please wait.

Presentation is loading. Please wait.

Introduction Using time property and location property from lost items’ pictures, we construct the Lost and Found System which combined with image search.

Similar presentations


Presentation on theme: "Introduction Using time property and location property from lost items’ pictures, we construct the Lost and Found System which combined with image search."— Presentation transcript:

1 Introduction Using time property and location property from lost items’ pictures, we construct the Lost and Found System which combined with image search technology. We find owners of lost items by similarity, computed in multiple dimension by using Skyline Algorithm, between items and owners. Due to the BigData century, our system is based on MapReduce architecture; having effective processing performance to reduce traditional human resource’s cost on finding items. In the poster, we will show details and operations of our system and display final results. Materials and methods (1)Nearest Neighbor algorithm NN algorithm will find the nearest one from the origin vertexes. After getting the point, it will ignore other points that can not replace the one we found. (2)Branch and Bound Skyline algorithm This algorithm stores all points into R-tree and use Heap to find out all dominant points. It solves the problem that NN algorithm will read and operate repeatedly in the multi-dimensional. (3)MapReduce Cloud architecture(fig. 1) which is proposed by Google is mainly used for massively data parallel computing. "Map" and "Reduce " are main concepts. Map is to do specified and independently operations to each element in set. Reduce is to merge a collection of elements appropriately and simplify its operation and results. Acknowledgments We thank George Chang for laboratory assistance and Phillip Chang for questionable statistical advice. Funding for this project was provided by the Ministry of Science and Technology and Data Management+ laboratory. Results After taking pictures of items that users found, users would upload pictures to the system. Our systems would run on Hadoop to get results. Figure 4. Found Interface Figure 5. Result Interface Table 1. Table 2. Conclusions By Installing our system’s APP version or using our system’s web version, users can create records for their items; user who find items can upload items’ pictures to database. After uploading lost items’ pictures, we use time property and location property from lost items’ pictures to compare with users’ items; system would find the likely users and notify them. Accurately analyze information, effectively reduce the time of processing, and correctly find items’ owners. 400410009 郭柏辰 400410063 許任傑 Department of Computer Science, National Chung Cheng University, Chiayi, Taiwan 19081 Literature cited Roussopoulos, N., Kelly, S., Vincent, F. Nearest Neighbor Queries, SIGMOD Rec., 1995. D. Papadias, Y. Tao, G. Fu, B. Seeger, An Optimal and Progressive Algorithm for Skyline Queries, In Proceedings of the 2003 ACM SIGMOD International Conference on Managementof Data, 2003. J. Dean and S. Ghemawat, MapReduce: Simplified Data Processing on Large Clusters,Communications of the ACM, 2008. A. Guttman, R-trees: A Dynamic Index Structure for Spatial Searching, In Proceedings of the 1984 ACM SIGMOD International Conference on Management of Data, 1984. For further information Please contact lf@ccu.edu.tw. More information on this and related projects can be obtained at dmplus.cs.ccu.edu.tw. Effective query of lost property base on skyline and map-reduce We use the following three properties to compute the result (1)Location We record users’ paths by GPS. If user finds an item and upload its picture, we can compute from users who went through these areas. The operation is like Fig. 2, the lost item’s location is at point a and User1, User2, and User3 are likely users. We find the nearest point on each path such as point a, b, and c in Fig. 2 by KNN algorithm. Figure 2. path (2)Time Using the time property of lost items’ pictures, we compute the difference between users’ and lost items’ time. In Fig.3, the time of finding lost item is Ta and the time of discovering the item is missing is Tb. Figure 3. time (3)Image Similarity After computing the first two properties, we use PerceptualDiff API to compute the similarity between pictures from users and pictures of lost items. Integrating the above description, we can get the follow formula : Ws + Wt + Wp = 1 (100%) (1) In formula (1), weight of location is Ws, weight of time is Wt, and weight of picture is Wp. Each weight means one dimension, and we use Skyline algorithm to compute the final result. data amount(K)execution time(sec) JAVA110.93 Hadoop115.26 data amount(K)execution time(sec) JAVA10015.66 Hadoop10015.37 data amount(K)execution time(sec) JAVA1000018.72 Hadoop1000015.53 Table 3. According to tables 1, 2, and 3, we found that the execution time of Hadoop did not increase much when data amount got bigger; on the contrary, the execution time of JAVA increase much. Figure 1. MapReduce Architecture Figure 6. Users would receive mails when their items was found


Download ppt "Introduction Using time property and location property from lost items’ pictures, we construct the Lost and Found System which combined with image search."

Similar presentations


Ads by Google