Research On Visual Perception Method Of Sensor Network For Water Environment Monitoring

With the continuous deepening of social industrialization and the development of urbanization,many water resources in our country have been polluted.Although our country has taken many measures to protect the water environment and achieved some good results,the situation of water environment is still serious,and more efforts are needed to be made to solve this problem.As a new and effective monitoring method,visual perception based on wireless video sensor network technology has been widely used in many monitoring area.Previous work of scholars and researchers has shown that water environment monitoring based on this technology is an important research direction.The network uses the reasonable sensor node deployment strategy to realize the full coverage of the monitoring area under the condition of relatively few sensor nodes and no monitoring blind spots.When target appears,the sensor nodes adaptively adjust their sensing directions to obtain the best monitoring perspectives.Then,the monitoring data is processed for target detection,and the salient regions are extracted from the background to complete the monitoring and identification process.Finally,the detection results are compressed and transmitted to the monitoring base station to realize the real-time and intuitive monitoring of water surface.This technology mainly includes the following three research contents: sensor node deployment and dynamic adjustment,image saliency detection and image compression transmission.This thesis studies the above three problems,the main research work is as follows:(1)A dynamic adjustment strategy of sensor nodes based on artificial fish swarm algorithm is proposed.The strategy uses fixed location deployment to place sensor nodes in the monitoring area,and achieves the complete coverage of the monitoring area in the case of relatively low coverage redundancy.After the deployment,the strategy uses the sensor monitoring direction preliminary adjustment strategy based on artificial fish swarm algorithm to adjust the monitoring directions of sensor nodes as a whole.By finding the local maximum coverage area of each node,the overall coverage area of network is maximized.Once the target appears,the sensor nodes regulate their sensor directions again to promote the monitoring area of the target and obtain the best monitoring visual perspectives.Then,among all the nodes participating in the target monitoring,the best node is selected according to their residual energies and detection effects.Experimental results show that the proposed strategy can not only obtain the greater coverage and better monitoring performance,but also balance the energy consumption of nodes and prolong the network life.(2)A saliency detection algorithm based on image sparse representation and combination of color coefficient is proposed.In this algorithm,an accurate background template is first generated through the background optimization selection strategy.According to the accurate background template,the algorithm uses the image sparse representation to get the primary saliency map.In the primary saliency map,the background regions can be well suppressed,but the saliency regions are not complete,and the differences between they and the background are small.Based on the primary saliency map,this algorithm adds a linear combination of multi-dimensional color coefficient strategy to obtain the improved saliency map.In the improved saliency map,salient regions can be highlighted completely.However,it contains some background false detection regions.At last,the algorithm combines the primary saliency map and the improved saliency map through the Bayesian fusion framework to get the final saliency map with the best detection quality.Experimental results show that the algorithm can effectively extract the salient region from the visible water image.(3)A saliency detection algorithm based on compressed reconstruction and linear combination of feature coefficients is proposed.The algorithm can comprehensively process the saliency detection of visible and infrared water images,which is more suitable for the needs of practical monitoring.It can adaptively choose the appropriate features according to the different categories of the input image,and effectively extract the salient region from the background.Firstly,the input image is processed by block based compressed sensing reconstruction technology to reduce the complexity of the method.Then,the algorithm uses the principle of feature contrast to roughly detect the salient region of the input image.Based on the primary saliency map,the proposed algorithm uses the best linear combination of feature coefficients to achieve more accurate results.Experimental results show that the algorithm is very effective for complex and changeable water images,and has good detection effects for both visible and infrared water images.(4)An adaptive image compression based on compressive sensing for video sensor nodes is proposed.In order to reduce the complexity of the algorithm and the storage space of the data,the input image is first divided into image blocks,and the subsequent processing is performed on each image block.Then,each image block is decomposed by the discrete wavelet for subsequently compression sampling.However,the coefficient matrix of each block is not suitable for compressed sampling because of its large size.A coefficient rearrangement is designed to reduce the size of coefficient matrix and measurement matrix.Next,the algorithm uses adaptive compression sampling strategy to compress each image block.The strategy adaptively allocates the appropriate sampling rate to each image block according to the amount of information of each image block,the residual energy of sensor node and the links quality of network.Through this strategy,the overall sampling rate of the image can be effectively controlled to balance the energy consumption of nodes and maintain the good quality of compressed image.Finally,the compressed data is packaged and transmitted to the host computer through wireless transmission.The data is restored to images through recovery algorithm and inverse wavelet transform on the host computer,realizing the monitoring data compressed transmission.By studying the deployment and dynamic adjustment of sensor node,image saliency detection and image compressed transmission,it provides theoretical support and technical solutions for improving the performance of water environment monitoring based on sensor network,and adds assistance to promote the application of this monitoring technology.