Xiaohua Tong

Neural network classification with optimization by genetic algorithms for remote sensing imagery

On the remote sensing imagery classification, the traditional methods based on statistical principle has the difficulties in distinguishing the objects with similar spectral characteristics, while the back propagation neural network method has the difficulties in sufficiency and convergence. Therefore, a new method based on neural network classification with optimization by genetic algorithms for remote sensing imagery is proposed in this paper. On the basis of the back propagation( BP )neural network classification, the optimization method by genetic algorithms is presented, including the numbers, the thresholds and the connection weights of nerve nodes of the hide layer in BP neural network. An approach on float coding with alterative length for genetic algorithms is proposed, and the evolution method is improved to obtain an optimal BP neural network. In the end, an experimental test on the remote sensing classification using TM image of Dianshan Lake is carried out, and higher classification accuracy has obtained compared to other methods, which is proved the feasibility and validity of the proposed approach.

Comparison of metaheuristic cellular automata models: A case study of dynamic land use simulation in the Yangtze River Delta

Cellular automata (CA) is a bottom-up modeling framework that has increasingly been applied to simulate land use change by capturing its dynamics. Metaheuristics such as particle swarm optimization (PSO), generalized simulated annealing (GSA) and genetic algorithm (GA) have widely been incorporated into CA modeling to generate more realistic simulation patterns. We present a comparative study of four CA models incorporating logistic regression (LR) and the three metaheuristics respectively to simulate land use change in the Yangtze River Delta from 2005 to 2015. The metaheuristic processes are guided by an objective function that represents the root-mean-square error (RMSE) of the transition rules, which can then automatically search for suboptimal CA coefficients. The three metaheuristics are substantially different in terms of the algorithm mechanism, optimization iteration, and computational time. The land conversion potentials from the metaheuristics are similar in global patterns but marginally different in local regions, which substantially differ from that calculated using LR. All three metaheuristic CA models simulated slightly less than the reference change while the LR-CA model simulated substantially more than the reference change, however all models allocated the change to similar places. Our study shows that the three metaheuristics can achieve similar outcomes in the optimization of CA transition rules and land use simulation, albeit with different sensitivities to their intrinsic control parameters. We suggest that any of the three metaheuristics could be used to construct land use CA models, if the algorithm complexity and computational time are not highly concerned.

Proving nonsingularity of coefficient matrix in least squares normal equation of non-linear semiparametric model

Non-linear Semiparametric model is a statistical model consisting of both parametric and nonparametric components, and the form of the parametric part is non-linear. The efficiency problem for a semiparametric model has been widely studied presently. Since non-linear parametric models have been studied deeply, and a set of basic theory have been set up, such as the measurement of the non-linearity of non-linear models and the statistics property of non-linear parametric estimation. Based on the nearest neighbor estimating theory of non-linear semiparametric models under the least squares principle, this paper proved the nonsingularity of coefficient matrix of normal equation under certain conditions. The nonsingularity of coefficient matrix of normal equation in least squares estimator of non-linear semiparametric models can be expanded to other least squares estimator of non-linear semiparametric models.

Weighted radial dimension: an improved fractal measurement for highway transportation networks distribution

An improved fractal measurement, the weighted radial dimension, is put forward for highway transportation networks distribution. The radial dimension (DL), originated from subway investigation in Stuttgart, is a fractal measurement for transportation systems under ideal assumption considering all the network lines to be homogeneous curves, ignoring the difference on spatial structure, quality and level, especially the highway networks. Considering these defects of radial dimension, an improved fractal measurement called weighted radial dimension (DWL) is introduced and the transportation system in Guangdong province is studied in detail using this novel method. Weighted radial dimensions are measured and calculated, and the spatial structure, intensity and connectivity of transportation networks are discussed in Guangdong province and the four sub-areas: the Pearl River Delta area, the East Costal area, the West Costal area and the Northern Guangdong area. In Guangdong province, the fractal spatial pattern characteristics of transportation system vary remarkably: it is the highest in the Pearl River Delta area, moderate in Costal area and lowest in the Northern Guangdong area. With the Pearl River Delta area as the centre, the weighted radial dimensions decrease with the distance increasing, while the decline level is smaller in the costal area and greater in the Northern Guangdong province. By analysis of the conic of highway density, it is recognized that the density decrease with the distance increasing from the calculation centre (Guangzhou), demonstrating the same trend as weighted radial dimensions shown. Evidently, the improved fractal measurement, weighted radial dimension, is an indictor describing the characteristics of highway transportation system more effectively and accurately.

A new methodology to mobile green spatial information service grid based on mobile agent

A new methodology of mobile green spatial information service is studied based on grid computing environment. Key technologies are presented in the methodology, including grid computing, mobile agent (MA) distributed computing, global positioning system (GPS), code division multiple access (CDMA), transfer control protocol/internet protocol (TCP/IP), and user datagram protocol (UDP). In order to deal with the narrow bandwidth and instability of the wireless internet, limited processing speed and low memory of mobile devices, a new mobile agent based mobile green spatial information service grid model is further proposed that has good load balance, high processing efficiency, less network communication and thus suitable for mobile green spatial information service. It can provide applications of green spatial information grid computing and mobile service. The theories and technologies architecture of mobile green spatial information service grid are built originally from the base, including green spatial information mobile agent model, distributed green grid geographic information system (GIS) server model, mobile agent server model and mobile green GIS client model. An application system for mobile green spatial information service grid is therefore developed authorship by visual c++ and embedded visual c++. A field test is carried out through this system in Shanghai, and the results show that the proposed model and methods are feasible and adaptable for mobile green spatial information service.

A new method integration of 3S for mobile landscaping survey and update

A new method integration of 3S (geographic information system (GIS), global positioning system (GPS), remote sensing(RS)) for mobile landscaping survey and update is studied in mobile GIS. Several Key technologies are presented in the method, including spatial data model on mobile personal digital assistant (PDA), GPS data difference processing and representation in GIS, GPS coordinate transformation model in mobile GIS, the method of PDA integration of Total Station, the RS image process based on Mobile Device. A new object-oriented spatial data model based on mobile GIS is proposed that has high query and display efficient. A new least squares-based GPS transformation model is further proposed that has fast calculation speed, high adjustment accuracy and thus suitable for mobile GIS. An integrative Application system integration of 3S for landscaping survey, update and management is therefore developed, including spatial data survey and update mobile GIS, spatial data processing and maintain desktop GIS and spatial data publication WEBGIS. A field test is carried out through this system in Shanghai, and the results show that the proposed model and methods are feasible for mobile landscaping information survey, update and gridding management.