In this paper,we propose a convolutional neural network（CNN）based on deep learning method for land cover classification of synthetic aperture radar（SAR）images.The proposed method consists of convolutional layers,pooling layers,a full connection layer and an output layer.The method acquires high-level abstractions for SAR data by using a hierarchical architecture composed of multiple non-linear transformations such as convolutions and poolings.The feature maps produced by convolutional layers are subsampled by pooling layers and then are converted into a feature vector by the full connection layer.The feature vector is then used by the output layer with softmax regression to perform land cover classification.The multi-layer method replaces hand-engineered features with backpropagation（BP）neural network algorithm for supervised feature learning,hierarchical feature extraction and land cover classification of SAR images.RADARSAT-2 ultra-fine beam high resolution HH-SAR images acquired in the rural urban fringe of the Greater Toronto Area（GTA）are selected for this study.The experiment results show that the accuracy of our classification method is about90%which is higher than that of nearest neighbor（NN）.
This paper proposes a wireframe model-based method for automated internal design. The method is used to extract geometric structure of an internal wireframe model and find out all loop structures of furniture models. The wireframe models are classified as the multiple independent sub-models according to the geometric structure by statistical analysis. The corresponding models are selected from a 3D model database to build an internal scene based on characteristic points of furniture wireframe models. In the experiments 3D database via manually selected 268 3D furniture models from Google 3D warehouse is built up. The experiments show that the method can construct 3D scenes in 1.1×103 ms. This method costs less time compared with traditional hierarchical method and depth-sensing camera method in the same experimental conditions. The method can be also used for 3D visualization either with complex backgrounds.
In order to solve the problem that the embedded software has the shortcoming of the platform dependence, this paper presents an embedded software analysis method based on the static structure model. Before control flow and data flow analysis, a lexical analysis/syntax analysis method with simplified grammar and sentence depth is designed to analyze the embedded software. The experiments use the open source code of smart meters as a case, and the artificial faults as the test objects, repeating 30 times. Compared with the popular static analyzing tools PC-Lint and Splint, the method can accurately orient 91% faults, which is between PC-Lint＇s 95% and Splint＇s 85%. The result indicates that the correct rate of our method is acceptable. Meanwhile, by removing the platform-dependent operation with simplified syntax analysis, our method is independent of development environment. It also shows that the method is applicable to the compiled C（including embedded software） program.
Software fault positioning is one of the most effective activities in program debugging. In this paper, we propose a model-based fault positioning method to detect the faults of embedded program without source code. The system takes the machine code of embedded software as input and translates the code into high-level language C with the software reverse engineering program. Then, the static analysis on the high-level program is taken to obtain a control flow graph（CFG）, which is denoted as a node-tree and each node is a basic block. According to the faults found by the field testing, we construct a fault model by extracting the features of the faulty code obtained by ranking the Ochiai coefficient of basic blocks. The model can be effectively used to locate the faults of the embedded program. Our method is evaluated on ST chips of the smart meter with the corresponding source code. The experiment shows that the proposed method has an effectiveness about 87% on the fault detection.
LIU JinshuoCHEN JianZHANG WeixinXU XiangyangYAN Jingjing
