To address the problems of low exploration efficiency and information exchange under limited communication bandwidth in the current Multiple Unmanned Aerial Vehicle （Multi-UAV） systems when exploring large-scale complex environments， a fast and fully autonomous exploration method for Multi-UAV in large-scale complex environments was proposed， including a fast and hierarchical exploration strategy and a lightweight large-scale environment modeling method. Firstly， closed viewpoints were generated in the front-end trajectory planning part to drive the Unmanned Aerial Vehicles （UAVs） to explore unknown environments. Then， the smooth， continuous， and time-optimal trajectory optimization problem was transformed into a convex optimization problem in the back-end， and this problem was modeled systematically. Meanwhile， in terms of environmental characterization， a random mapping method was used for lightweight mapping and map data interaction. Finally， in simulation， the proposed method was compared with fast exploration method using incremental boundary information and hierarchical planning — FUEL （Fast Unmanned aerial vehicle ExpLoration）， rapid exploration method based on frontiers — FBE （Frontier-Based Exploration）， and exploration method based on the next best viewpoint — NBVP （Next Best View Planner）. The results show that the proposed method improves the exploration time performance by 14.4%， 43.9% and 47.7%， respectively， and the lightweight mapping method reduces the data size by 28.3% and 22.4%， respectively， compared to the Bayesian method and the polyhedron method. It can be seen that the proposed method can perform fast and fully autonomous exploration in large-scale complex environments efficiently.

To meet the requirements for adaptivity, and robustness of the algorithm to optimize urban traffic signal control, a traffic signal control algorithm based on Deep Reinforcement Learning (DRL) was proposed to control the whole regional traffic with a control Agent contructed by a deep learning network. Firstly, the Agent predicted the best possible traffic control strategy for the current state by observing continously the state of the traffic environment with an abstract representation of a location matrix and a speed matrix, because the matrix representation method can effectively abstract vital information and reduce redundant information about the traffic environment. Then, based on the impact of the strategy selected on the traffic environment, a reinforcement learning algorithm was employed to correct the intrinsic parameters of the Agent constantly in order to maximize the global speed in a period of time. Finally, after several iterations, the Agent learned how to effectively control the traffic.The experiments in the traffic simulation software Vissim show that compared with other algorithms based on DRL, the proposed algorithm is superior in average global speed, average queue length and stability; the average global speed increases 9% and the average queue length decreases 13.4% compared to the baseline. The experimental results verify that the proposed algorithm can adapt to complex and dynamically changing traffic environment.

A data communication system based on High-level Data Link Control (HDLC) protocol was designed, and then HDLC protocol in Field Programmable Gate Array (FPGA) was implemented. The system can take advantage of FPGA hardware effectively. It has some outstanding advantages, such as high scale integration, no need of peripheral circuit and being easy to use. This paper emphatically described the implementation methods of transmitting protocol and receiving protocol and the method for "0" bit insertion and deletion.

A module of object detection based on Java and OpenCV was presented, every implementation step of the module's block diagram was explained in detail, the specifies of using Java Native Interface （JNI） to implement the function of object detection via OpenCV were mentioned. This module could simply integrate most academia and industry video based application systems. The experimental results show that the Java video system which integrated this module has a high detection rate and realizes real-time processing.

Fingerprint image quality estimation is an important step of Automatic Fingerprint Identify System (AFIS). The quality of fingerprint image from fingerprint machine seriously affects preprocess and identification. The exiting detailed regulations were analyzed, the influence on fingerprint quality from valid area, range of gray level, dry, wet to deflect were discussed. A new method was proposed to estimate fingerprint image quality based on information about edge-minutiaes, and qualified fingerprint images were provided. Experimental results show that it can filter unqualified fingerprint image effectively, and it is helpful to raise the efficiency of whole system.

The wireless sensor network is event triggering network, which contains many nodes. The concepts of same source collision and different source collision were put forward. The collision caused by identical event triggering several nodes movement was regarded as same source collision and the collision caused by different events triggering different nodes movement was regarded as different source collision. All these two collisions can cause the decline of network data stream and the waste of nodes energy. Therefore, it is necessary to realize collision avoidance for MAC layer based on existing MAC protocol of wireless sensor. The collision aware MAC (CAMAC) using the idea of filter and power control was proposed. The principle of CAMAC handling different collisions was discussed, and CAMAC was simulated by NS.

A GroundDelay Policy (GDP)model with its algorithm with constrained airpor's arriving capacity was analyzed and modified. Different from conventional models, the Discrete Event System (DES) method was adopted to analyze the air traffic flow management. The simulation result shows that compared with the conventional GDP model based on linear planning, the GDP model and algorithm with constrained airport's arriving capacity can not only solve the continuity of flight time efficiently, enhance the precision of optimized arriving time, but also reduce the complexity of total calculation significantly.

Gaussian descriptors are contourbased shape features. They are invariant to translation, scaling changes, rotation and reflection. Compared to the existing shape features, they are more robust against noise and slight edge variations, and have lower computation complexity and higher recognition/retrieval rate. In addition, they are applicationindependent. In this paper, Gaussian descriptors were used as features for character recognition. A comparison with another contourbased moment invariants, which is an improvement and extension of classical Hu moments, was also given. Numerical experimental results show that Gaussian descriptors are an attractive tool for character recognition.