**7. Acknowledgment**

The authors would like to thank Tadaaki Shimizu, Yusuke Hamada, Naoki Ishibashi, Shinya Iwasaki, Hirotoshi Okumura, Masato Hamamura, and Shingo Iiyama in Tottori University.

### **8. References**

36 Recent Developments in Video Surveillance

extracted H-HVS color. A solid line rectangle encloses the part where a red target was discovered, the rectangle of the short dashed line encloses the part where a blue target was discovered, and a dotted line rectangle encloses the part where a green target was discovered. The plot becomes a constant pattern if target is captured correctly, because the movement of the target is periodic. However, The illuminance was actually different according to the place that the camera installed, the color was judged the black by the shadow, and the target might not be able to be caught. Therefore, the number of agents did not increase and decrease by a constant pattern. But, the Neighbor node determination algorithm was able to calculate neighbor camera nodes. However, when the target is captured by video camera, the neighbor nodes are recomputed by the algorithm, then it was confirmed that an agent might not be deployed appropriately for a few dozens milliseconds. About this result, it is necessary to improve the computation time. In graphs of Fig. 10, the plot shows the time the target captured. It is confirmed that a target is lost temporarily from the camera capturing the target. But it is confirmed that the human tracking keeps tracking targets continuously by capturing the target by other cameras.

A construction method of automatic human tracking system with mobile agent technology is proposed using neighbor node determination algorithm. The construction method consists of the neighbor node determination algorithm and tracking methods. The neighbor node determination algorithm can compute neighbor nodes. This algorithm can be efficient to compute neighbor node and can make the system robust even if view distance of camera is changed. The tracking methods consist of follower method and detection method. The follower method can identify feature of a target locally. The detection method can search a lost target but a search cycle has to be within *walking speed × distance between cameras*. The detection method can be efficient to detect a target if the search cycle is near the walking speed. A mobile agent can keep tracking a target by using these detection methods if the agent lost the target. In addition, from the experiment results, the Stationary net detection method can detect a target faster than the Ripple detection method. And the Stationary net detection method can use smaller number of agents than the Ripple detection method. Because the Ripple detection method searches a target by widening a search gradually but the Stationary net detection method can widen a search efficiently by the Neighbor node

The effectiveness of proposed tracking method was experimented using simulator and real environment. In the experiment using simulator, the tracking methods are experimented by the walking speed of a target. In the detection methods, consideration is added about the propriety of the parameter *n* which gives the number of non-camera nodes. Aimed to confirming behavior of the automatic human tracking system, the system in a real environment uses the simple image processing which can identify the color information of a target except the influence by the accuracy of image processing. And the follower method and the detection method are confirmed to be effective by a toy instead of a targeted person

We will research more efficient detection to improve the automatic human tracking system. In addition, the accuracy of image processing has to be improved more to track a target more accurately. We are considering to improve our tracking system by combining effective

and to be able to construct the automatic human tracking system.

**6. Conclusion** 

determination algorithm.


**3** 

*1,3Vietnam 2France* 

**Appearance-Based Retrieval for Tracked** 

*3IFI, MSI Team; IRD, UMI 209 UMMISCO; Vietnam National University,* 

Video surveillance is a rapidly growing industry. Driven by low-hardware costs, heightened security fears and increased capabilities, video surveillance equipment is being deployed more widely and with greater storage than ever. This provides a huge amount of video data. Associating to these video data, retrieval facilities become very useful for many purposes and many kinds of staff. Recently, several approaches have been dedicated to retrieval facilities for surveillance data (Le, Thonnat et al. 2009) (Zhang, Chen et al. 2009). Figure 1 shows how indexing and retrieval facility can be integrated in a surveillance system. Videos coming from cameras will be interpreted by the video analysis module. There are two modes for using the analysed results: (1) the corresponding alarms are sent to members of the security staff to inform them about the situation; (2) the analysed results are stored in order to be used in the future. In this chapter, we focus on analysing current achievements in surveillance video indexing and retrieval. Video analysis (Senior 2009) is beyond the

Video analysis module provides two main result types of result: objects and events. Thus, surveillance video indexing and retrieval approaches can divided into two categories: surveillance video indexing and retrieval at the object level (Calderara, Cucchiara et al. 2006; Ma and Cohen 2007; Le, Thonnat et al. 2009) and at the event level (Zhang, Chen et al. 2009; Velipasalar, Brown et al. 2010). As events of interest may vary significantly among different applications and users, this chapter focuses on presenting the work done for

The remaining of the chapter is organized as follows: In Section 2, we give a brief overview of surveillance object retrieval. Section 3 aims at analysing in detail appearance-based surveillance object retrieval. We first give some definitions and point out the existing challenges. Then, we describe the solutions proposed for two important tasks: object signature building and object matching in order to overcome these challenges. Section 4

presents current achievements and discusses about open problems in this domain.

surveillance video indexing and retrieval at the object level.

**1. Introduction**

scope of this chapter.

**Objects in Surveillance Videos** 

*2PULSAR, INRIA Sophia Antipolis,* 

Thi-Lan Le1, Monique Thonnat2 and Alain Boucher3 *1MICA Center, HUST - CNRS/UMI 2954 - Grenoble INP, Hanoi,* 

