Key Technologies Of Intelligent Image Processing For Urban Transportation

Internet of Things(IOT)is the 3rd development wave of word information domain,which has the huge good future as a hot technology.IOT is one of key technologies to support intelligent city/urban transportation.Under the banner of the intelligent city/urban transportation based on IOT,the image intelligent processing is one of important aspects components.By improving infrastructure,implementing special urban traffic lanes,rapid urban traffic,rail transit,and with the application of advanced information technology,urban traffic has developed rapidly and played an important role in urban residents' travel.Based on establishing and improving the traffic image database,which is big data,from the perspective of system engineering,on the basis of analyzing the influencing factors of traffic planning and the investigation and analysis of urban traffic capacity and operation status quo,the establishment of urban intelligent traffic dispatching system is of great significance to the construction of intelligent city.By analyzing the function of image denoising,image coding and image integrating,we have studied the relative novel methods to solve the problem of intelligent city/urban transportation.In this dissertation,the main innovative contribution of the research work is as follows:We proposed a kind of new method of compressed sensing image acquisition and intelligent image denoising based on compressed sensing.The random matrix which is generated by the seed vector to get from the diagonal matrix is created by applying Bernoulli matrix.Toeplitz matrix is used as basic comparison method,our designed new matrix has obvious advantages of variables and the information of matrix.A new non-linear adaptive shrinkage function is designed,which has some advantages of fast approaching true wavelet coefficient value to get the relative features of the images,such as edge and contour.Meanwhile,we design the multi-scale model product method to easily extract the image information and delete the noise information.We proposed a kind of novel method of intelligent image coding based on self-adaptive transmission.To meet the load balancing and quality of service requirements of city/urban transportation,we designed the cross-layer optimization strategy to control the routing metric and protocol of self-adaptive transmission after considering the drawbacks of AODV.New the EAODV protocol is designed which is better than the original AODV in data packet delivery ratio,end-to-end delay,normalized routing overhead and routing.At the same time,a kind of new method of intelligent image coding has been designed,which coding by using the coefficients on the middle row and middle column to predict the un-kown wavelet coefficients for low-frequency part,but we deal with the high-frequency part using new coding strategy.During the coding,we use the method of matrix of maximum pixel to reduce the redundant coefficient scans,so the higher compression ratio can be got.We proposed a new method of intelligent image integrating based on the learning strategy of fuzzy-neural network.The method has been composed the advantage of fuzzy theory and neural network in this method,because the neural network can process fusion in parallel paradigm for image and pattern identification,the fuzzy theory can make full use of expert knowledge and efficient understanding for un-exact data.The method clusters the image by self-learning and self-adaptive capability firstly,then makes the result into fuzzy process,and gets membership vector for each pixel of image to fuse the image.The method is very useful for fusion of multi-source image.When the integrating of the image is done,the method weighs the similarity of the color information and gradient construction.The colors and gradients is as a measure of the similarity of between the two matching image patches,the extent of the similarity is improved.By comparing the similarity of histograms of the image blocks and histograms of input sample image blocks,the method can get the best size of image block based on the the learning strategy of fuzzy-neural network.Based on our experimental tests,the above presented methods have been verified.We use the new MATLAB tool kits to test our aforementioned methods.The comparison experiments have been done.The testing results have shown that our methods have good performances.The effects of our methods have been verified in the application of intelligent city/urban transportation based on IOT.