Aiming at the problem of low detection rate of depth feature based object detection method Faster R-CNN （Faster Region-based Convolutional Neural Network） in flame detection tasks, a color-guided anchoring strategy was proposed. In this strategy, a flame color model was designed to limit the generation of anchors, which means the flame color was used to limit the generation locations of the anchors, thereby reducing the number of initial anchors and improving the computational efficiency. To further improve the computational efficiency of the network, the masked convolution was used to replace the original convolution layer in the region proposal network. Experiments were conducted on BoWFire and Corsician datasets to verify the detection performance of the proposed method. The experimental results show that the proposed method improves detection speed by 10.1% compared to the original Faster R-CNN, has the F-measure of flame detection of 0.87 on BoWFire, and has the accuracy reached 99.33% on Corsician.The proposed method can improve the efficiency of flame detection and can accurately detect flames in images.

Concerning that using camera and sensor to detect human motion has the shortcomings of difficult deployment, expensive device and blind zone, a method of human motion detection using Wireless Fidelity (WiFi) signal was proposed. Firstly, the wireless network card was used to receive the Channel State Information (CSI) of the WiFi in the detected environment. Secondly, the Local Outlier Factor (LOF) algorithm and the Hampel filter were used to remove the abnormal CSI data. After the frequency shift caused by the rough synchronization of the network card clock was removed by the linear regression algorithm, the Principal Component Analysis (PCA) was used to reduce dimension and Naive Bayes algorithm was used to classify the CSI data in different cases, which generated a model for judging human movement states. Finally, the model was used to judge the state of human motion. In the experiment, the proposed method can quickly determine the state of human motion and reach the correct rate of 95.62%. The experimental results show that the proposed method can detect and identify the movement of people well.

To solve the excessive authorization problem of Android, this paper proposed a User-Friendly privacy monitoring and management Mechanism on AnDroid named UFMDroid. Proxy redirect technology was used to implement privacy-related behavior monitoring module and fine-grained resources constraint module in Android control flow. UFMDroid analyzed the existing applications on Android market and constructed a permission profile as preset by hierarchical clustering and Euclidean distance metric to filter suspicious authority. A static threat value could be provided by calculating the distance between the preset and the current permission configuration. The privacy-related behaviors of application were classified and both of the individual threat and combination threat were considered in calculating the dynamic runtime threat value. In addition, fake data mechanism was imported to prevent the application from crashing while the permission was withdrawn. The experimental results show that UFMDroid can monitor the usage of 21 different resources and it can intercept the privacy leakage behaviors in accordance with user configuration. UFMDroid can enhance the security of Android to some extent.

The definition of the CORBA component model with IDL3 and CIDL was analyzed, and the compiler and its interface repository of the CORBA component model were designed. The front end of compiler was implemented by analyzing accidence and syntax parsing on source file with LEX and YACC, while the back end was implemented by "preorder traversal" means on interface repository, thus providing generate tools for CORBA component platform to make components.