Research And Implementation Of Audio Collection And Image Search System In Multimedia Sensor Network

As a novel information acquiring and processing technology, compared to the traditional sensor networks, people pay more attention to the information-intensive data(e.g. audio, video, image). Potential applications of multimedia sensor networks span a wide spectrum from military to industrial, from commercial to environmental monitoring. This study focuses on how to realize the high-speed audio data acquisition compression and reliable data transmission on resources limited sensor node, and speeding up image processing in network. The main works are as follows:(1) We described an audio data acquisition protocol. Different from the acquisition of environmental data such as temperature and light,audio data collection requires high frequency sampling, and will produce a large number of data. Therefore, posing new challenges to data collection, storage and transmission in resource restricted wireless sensor network. Based on the above-said characteristics, an audio data collection protocol with storage, compression and reliable transmission was proposed. A prototype system based on the protocol was implemented on MicaZ nodes. The results of experiments show that the prototype system can provide reliable transmission and reduce the number of data.(2) We designed a distributed image search system. The SIFT detector and descriptor is a common object recognition algorithm in image processing. However, it takes a long time to run the SIFT algorithm on resource restricted sensor nodes. In this paper, we design a distributed SIFT algorithm and put forward RCT image compression algorithm. Analysis and experimental results show that the distributed SIFT algorithm can accelerate the search speed while the RCT algorithm needs less computation which can effectively save node energy.(3) Based on the above research work, we implement an audio data collection system and an image search system on Micaz and Imote2 nodes which demonstrate the performance of the protocol proposed in the paper. We use ADC components to collect audio data in high-speed and join retransmission mechanism and compression algorithm in data processing layer. We can uncompress data and analysis protocol performance in monitoring layer. We modprobe tos_mac module in Imote2 nodes to enable wireless communication. It uses SIFT algorithm to realize object matching. Experimental results show that the audio data collection system is stable and the image search system which can find the right object when image translation, rotation and affine occurs is fast.