TMCnet News
Redundancy Data Elimination Scheme Based on Stitching Technique in Image Senor Networks [Sensors & Transducers (Canada)](Sensors & Transducers (Canada) Via Acquire Media NewsEdge) Abstract: Using image stitching technology to find out overlap areas in the camera node capture image, inform the camera node, camera node will filter redundancy image data which is overlap with adjacent nodes, and then cluster nodes rebuild images which has been received, stitch together, form a panorama. Through this process, it can reduce image redundancy data affectivity in densely deployed wireless camera image sensor network, the network life cycle can be greatly improved. This paper proposes a new calculation method. Copyright © 2014 IFSA Publishing, S.L. Keywords: Image sensor networks, Suture technique, Redundant data. (ProQuest: ... denotes formulae omitted.) 1. Introduction With the development and social economic progress, especially in science and technology constantly updated, the development of microprocessor technology and sensor technology is faster and faster. In support of these advanced technologies, we can induce sensor networks effectively capture a variety of multimedia information, such as audio, video, images, etc., and gradually produced a wireless multimedia sensor network [1]. With the practice and study, we found that wireless multimedia sensor networks node is often only a very short life cycle, it is because wireless multimedia sensor networks need to collect a lot of information, and require a lot of computation, which would greatly shortens the survival of node cycles. This problem has aroused people's attention in wireless multimedia sensor networks, using processors and sensors, and wireless communication module only low power consumption, the use of a battery with a larger capacity. So in this case, a very urgent problem is to take a series of measures to achieve battery life and improved network lifetime purpose. Currently, research in this area, there has been a lot of researchers have conducted some research and analysis, a series of measures to achieve energy efficiency while increasing the lifetime of network nodes. In this paper, on the basis of these studies, the innovation and development, we propose a new algorithm, the image stitching technology applied to the image sensor networks, the overlapped portion of the image to clear, so the amount of data will be effectively reduced, communications naturally lower energy consumption and prolong network lifetime. 2. Research Background Wireless Sensor in the data collection process, there are many cases will result in the generation of redundant data, first redundant data exists in the same multimedia sensor node collected at the same time between successive images. Second, it is adjacent to the multimedia sensor node at the same time to the same perception interest area of monitoring, there are redundant data between the collected images. Thirdly, it is redundant data exists in the multimedia sensor nodes collect between adjacent pixels in an image. To solve these problems, we need to take a series of prompting, to eliminate the redundant data, or if the aggregation node to send multiple multimedia sensor nodes to collect all the data, will be on a serious waste of communication bandwidth, and because of the need transfer large amounts of data, the entire network will consume too much energy, the life cycle of the network have an adverse impact on the scale of network applications adversely affected. For this problem, the image stitching techniques can be used to resolve. Image stitching technology is a wireless multimedia sensor networks is very important to a data processing technique. Through practice studies have shown that the application of wireless multimedia sensor image stitching technology that can handle from a node in the network to obtain a variety of raw data, which will be part of the complex to the active redundancy eliminated, the network data traffic has been effectively reduced energy consumption of this node can be effectively reduced, the life cycle of the network can also play a role in prolonged. 3. Concept of Data Fusion Because there are relatively dense node deployment in wireless multimedia sensor networks, so if multimedia sensor has a nearby location, the information collected may be repeated and redundant. If you want to transfer each node aggregation layer data collected, it will waste a lot of communication bandwidth. Also, large amounts of data, but also the amount of energy consumed by the network, the network lifetime effects. For this problem, data fusion techniques can be used, it refers to the multimedia information within the network collection process, the use of multimedia sensor node's local computing and storage capacity to handle this information collected, that repeat redundant information to eliminate, so that you do not need those unnecessary data transmission, the nodes can be effectively transmitted energy savings. In addition, comparative analysis of multimedia sensor nodes can collect information that will individual nodes to which there is a large collection of information to the exclusion of error, so that you can improve the reliability of monitoring information. 4. The Advantages of Data Fusion Technology in the Application of Image Sensor Network Through practice research shows that the image sensor to eliminate redundant data network aspects of using suture techniques has great advantages, embodied in three aspects: First, the network can be a large degree of energy savings: Through the above analysis, we have learned that the narrative, because the wireless multimedia sensor networks is relatively dense deployment node, then if multimedia sensor node has an adjacent location, collected data may be repeated and redundant; if all these collected data to the sink node, the node will be a great waste of energy, and therefore need to deal with the redundant data. The suture techniques for data fusion technology is an important one, can be effectively applied to the image sensor to eliminate redundant data among the network, the above -mentioned algorithm can be used to eliminate the image sensor network nodes redundant data collected, such amount of data needs to be transferred has been greatly reduced, multimedia sensor nodes has been effectively transmit energy savings, and thus the entire network for effective energy savings and prolong the network lifetime. Second, the accuracy of data collection can be enhanced: Because usually deployed wireless multimedia sensor network environment is unattended, then the harsh environment on the node itself may affect the normal operation of features, some exceptions, report an error data. Thus, single multimedia sensor nodes collect data on the possible existence of errors, and the use of data fusion technology can be the same object multiple multimedia nodes collect data for monitoring and analysis of the results, the error data is excluded, thereby increasing data accuracy and reliability. Third, the data collection efficiency can be significantly improved: we have already mentioned above, if the multimedia sensor node has a nearby location, then there will be a lot of redundancy exists in the data collected on these unnecessary the data to be transmitted, the communication bandwidth will result in serious waste, and the network will be limited additional energy consumption. Thus, the need to use the data fusion technique to the amount of data to be transmitted reduced, so that the burden of network traffic can be greatly reduced, network traffic efficiency can be improved, the transmission delay has been greatly reduced, to achieve data collection the purpose of improving efficiency. 5. Overview of Image Stitching Technology Image stitching technology as a new technology, referring to the stitching multiple images, making it a whole lot of panoramas, these images are the existence of spatial correlation. Then it can be in the wireless multimedia sensor networks application of this technology, in order to set the camera node transmits those to eliminate redundant image data, serve to reduce transmission energy consumption purposes [1]. Literature [2] for the first time in wireless multimedia sensor networks introduces image stitching technology, simulation results show that the introduction of suture technique over the image, the image sensor has a relatively dense cover, and there are a lot of image overlap condition exists, so the transmitted data the amount can be effectively reduced, the transmission power consumption can be greatly saved, but it will significantly increase the data in the calculation of energy consumption, and most calculations are repeated. Based on the literature [1], based on the use of automatic image stitching method to describe the spatial correlation of the image, and by means of automatic image stitching method for image data redundancy greatly reduced, so that the amount of data transmission is greatly the reduction, the transmission power is reduced; on this basis, proposed a new method of calculating the object of this method of calculation of the correlation coefficient of the image. The specific method is the case, defined the image pixel as (Xi, Yi), which is collected camera nodes, that the number of it is 'M\ Because the camera sensor network nodes use the same terms in the model, without loss of generality, then we can set those images' resolution as 'M' rows and 'N' columns [3]. After space conversion, the original coordinates (Xi, Yi) will be changed to (Xy, Yy). Number 'i' image space transformation matrix is defined as Tik, k has a value of 1, 2, 3, 4, then after transformation, new pixel coordinates (Xy, Yy) can be expressed as: ... (1) In this formula, it is possible to draw translation transform, respectively, rotation transformation, scaling transform and shear transformation matrix, respectively with Tii, Ta, Ti3 and Ti4 to represent [4], 6. Methods Description Random dense deployments camera nodes that do not optimize it, then there must be a certain degree of redundant nodes by the camera image data were collected. So in order for the camera to reduce the energy consumption of node communication, a very effective way is to eliminate redundant data exists in this part, to ensure that the next hop node transmission, will not have this part of the redundant data. Another proposed in the literature to remove redundant data, this method proposed is based on 2D camera node coverage model. This method is specifically referring to is the use of the camera's coverage area overlaps the node position, combined with its relations perspective, will this type of camera images collected node redundant part to find out, so that the camera nodes only part of the data independently transmission, the adjacent base station node images can be spliced [5]. In this way, if the camera is at the coverage area of the node is the same horizontal plane, then the transmission of redundant data can be greatly reduced, however, we need to note that, due to the various aspects of the constraints, the actual image sensor networks are often unable to meet these requirements [6]. In this paper, hypothesized a network topology, the topology is divided into clusters; as shown in Fig. 1: In the network topology, the camera node cluster head nodes to the cluster management, computing capability of cluster head nodes meet the relevant requirements, and the storage space and battery capacity. We use the technique of suture to handle the camera node images within the cluster, and assuming that the network has been in accordance with the clustering method to cluster was good right division. Based on this idea, we put forward a new method: in the beginning of the work, the image to the cluster head node sends the camera node in this cluster are collected; the camera image processing nodes is accomplished by a cluster head node. It contains these contents. One is to change the first acquisition or camera node position and angle, the image then can be combined with the reference to the method of image stitching method of three steps to handle M camera node to collect [7], and in the process saving process, spatial variation of the image matrix. Two without changing the camera node position and angle, then in the image transformation and splicing of M camera node collection, only need to use the original parameters can be calculated. On this basis, the redundant data on camera node transmission to reduce. After image stitching, we will set each sub graph' parameter as Uu, Vit, Xu, Yu, which is in corresponding to 'UíT,'VíT,'XíT,'Yii', so parameter is Un=u, Vn=v, Xü=XíiRí, Yii= YuRi. Need to pay attention to is, we talked about here at Uu, Vu, Xu, Yu number of lines and the pixel position, are relative, relative to a reference position of each sub graph itself, the reference position new image it is not generated after the suture. Image stitching is good, then the corresponding to each part of the original image is called sub graph. We denoted the scaling factor of 'I' image stitching process as 'Ri' [8]. The cluster head node send Xu and Yu to the number T camera node, so in the next acquisition, camera node only needs to combine the values Xu and Yu, and from their pixel send a position to the cluster head which collected from CMOS image sensors; the camera to collect the cluster head node, the image has no transmission of partially filled, then rectangular image reconstruction [9], and send the camera node, so space transform the next and splicing work will become easier. Through selected by the image to the base station, nodes send primary suture of the cluster head node, the base station node to apply image panorama strengthen, gain adjustment to primary suture, image bandwidth range and to further optimize the image after primary suture, the visual effect of increased. 7. Calculation Method for Correlation Coefficient of Image If the camera nodes are densely deployed, the coverage area overlap phenomenon is bound to exist, can be measured by the correlation coefficient the degree of overlap of these regions. In the references given in the calculations to cover the camera node 2D model to calculate the correlation coefficient and other references are also given in the method of calculation, in order to cover the camera model 3D node correlation coefficient calculation. Both methods, however there are significant limitations that these two calculation methods are based on spatial location of the camera, that camera node position and perspective is necessary to understand, so that the flexibility of the camera node deployment will be greatly limited role in reducing the extent of its fault tolerance, and requires the camera has a sensor node, the only way to take their own position and angle of perception, but the node cost and power consumption has been greatly node increase [10]. The proposed method of calculating the actual face of the camera image acquisition correlation coefficient, by this method may be that the correlation coefficient based on the spatial location of the node calculation of the problems solved effectively. We will defined number 'i' and 'j' image' Correlation Coefficient as 'Py', and so as to satisfy Py = Cy / Sy, Py and meet between 0 and 1. The overlapping area of pictures T and image 'J' is expressed as 'Cy', and the total area of pictures T and image 'J' after spliced is expressed as 'Sy'. The number of pixels used to represent the area. Then Py can be calculated using the following method: 1. If there is bonded relationship exists between the i and j, then turn to 2 directly, or can make Py equals to 0. 2. If the image from left to splice 'j' and 'i' image, then you can come to such a formula: ... (2) 3. If the image from right to splice 'j' and 'i' image, then you can come to such a formula: ... (3) What combination of these images can be defined in the correlation coefficient with a correlation coefficient of two images to lead out. We make such a definition, o= Ca/Sa, the collection of images to examine represented by 'a', all the overlapping regions in image 'a' is expressed as 'Ca', total image area of all individual images in image 'a' is expressed as 'Sa', the image area which does not overlap each other separate part in picture 'a' is expressed as 'Da', then you can come to such a formula: ... (4) ... (5) 8. Simulation Combined with the steps mentioned above, we toke a simulation. We will focus on co-existence of a group of images as experimental subjects, as shown below (Fig. 2), these three test images have the same size. After the simulation, we found that the proposed method has good adaptability, can effectively solve the problem of data redundancy. 9. Conclusions This paper propose a concept and calculation methods about correlation coefficient of the camera image, use the image stitching technology for wireless multimedia network camera nodes reduce redundant data, the correlation coefficient calculated on the original issues that arise from the fundamental to be solution. The results show that using the proposed method, the redundant data transmission can be effectively reduced, the camera node communication consumption has been reduced. References [1]. S. Zhenghui Ma, Ruzhuang Wang, Haiping Huang, Image sensor networks based on the technique of suture to eliminate redundant data, Journal of Southeast University, 2, 10,2012, pp. 123-124. [2]. S. Zhenghui Ma, Wireless multimedia sensor network image stitching technology research, Nanjing Post and Communications University, 2, 1, 2012, pp. 43-45. [3]. S. Rong Wu, Wireless multimedia sensor network data fusion technology research, Nanjing Post and Communications University, 5,2, 1995, pp. 10-15. [4]. S. Zheyuan Xiong, Wireless multimedia sensor network image coding algorithms, Central South University, 2, 5,2012, pp. 43-45. [5]. S. Lu Tao, Zhu Qing-Xin, Zhu Yu-Yu, An EnergyEfficient Adaptive Clustering Protocol for Wireless Sensor Network, Sensors & Transducers, Vol. 152, May 5, 2013, pp. 41-50. [6]. S. Lili Wei, Lightweight wireless sensor network security research of data fusion scheme, Nanjing Post and Communications University, 2, 3, 2012, pp. 99-100. [7]. S. Zhi Zhong, With mobile nodes of wireless sensor network localization algorithm and data collection protocol research, Nanjing Post and Communications University, 5,2,2012, pp. 67-69. [8]. S. Zhiwei Lee, Based on the clustering of the wireless sensor network lifetime extension strategy research, Ocean University of China, 2, 5, 2006, pp. 54-57. [9]. S. Lei Yan, Xiaokang Ding, Zheng Yu, Jianlei Kong, Jinhao Liu, A Novel Identification Method of Obstacles Based on Multi-sensor Data Fusion in Forest, Sensors & Transducers, Vol. 155, Issue 8, August 2013, pp. 39-46. [10]. Hua Ouyang, Hui Li, Mei Qian. Compression of Power Quality Data Based on Improved DCT Transform, Sensors & Transducers, Vol. 19, Special Issue, 2013, pp. 13-18. Chunling Tang Chongqing Technology and Business Institute, No. 1, Hualong Avenue, Jiulong Science and Technology Park, Jiulongpo District, Chongqing City, 400052, China Tel: +86 13883331650, fax: 13883331650 E-mail: [email protected] Received: 21 March 2014 /Accepted: 30 April 2014 /Published: 30 June 2014 (c) 2014 IFSA Publishing, S.L. |