Search

In view of the ghost and time cost of the Vibe, an improved Vibe algorithm based on Difference of Gaussian (DoG) is proposed in this paper. Firstly, video frames are processed by Gauss Pyramid, the original layer, the first layer and the second layer of three images are obtained. Secondly, three-frame difference is performed and fused, then rough moving target is yielded. Thirdly, the target pixel is judged based on the improved vibe algorithm and the complete moving target is achieved. The improvement on the traditional Vibe algorithm is to enlarge the range of sample pixels selection, avoiding repeated selection of error samples. On the other hand, the background of DoG images is updated and moving objects is segmented. Finally, the result is fused by three-frame difference and the improved Vibe algorithm. The experiment results show that ghost phenomenon can be restrained efficiently by the proposed algorithm and the calculation speed is also improved. The feature of real time can also be ensured.

France and China both are of typical power centralization with similar political management systems, and there are same and different aspects existing in higher education. so comparison between the two countries is meaningful. Curriculum provision, innovation ability cultivating and training mode are compared between France and China. Development trend of graduate education is also presented. Some enlightenment may be obtained to improve Chinese higher education level further.

In fire protection field smoke recognition is becoming more and more significant compared with traditional method. Vision-based smoke detection approach using image processing technology in smart monitoring system is proposed for early alarm. Firstly, to improve the image quality, fuzzy logic is employed for image enhancement. Then candidate Smoke area is extracted by Gaussian Mixed Model. According to the features of smoke, static and dynamic features are extracted. Thirdly, hyper parameter of Support Vector Machine(SVM) is determined and then the model of SVM is established. Finally, the characteristics vectors are input into SVM for smoke recognition online. Experiments are made by several videos under a variety of conditions and it is shown that this proposed algorithm is feasible and effective. It also has strong robustness and high real time.

In high space fire is one of the great threats for human security. The previous temperature-sensed and smoke-sensed method can't respond quickly to early fire, and the fire may not be detected immediately if it is far away from the sensors. Therefore, double band infrared vision-based fire detection is adopted to overcome the drawbacks of traditional detection equipments for high space. Firstly, a region partition algorithm is proposed, which improves the recognition efficiency. The flame features are extracted and normalized, and then a BP neural network model is established for recognition combined with image processing. The experimental results demonstrate that the proposed method can distinguish fire effectively and improve the accuracy. In addition, the false alarms rate issued to traditional approach can be reduced drastically. This approach has short response time and strong robustness.

Fire detection is very important in safety area, and at all times it is a difficult problem in coal mine. In this paper, a real-time fire-detector algorithm is proposed. In early stage there is no evident flame, and the temperature is low, the visible light is scarce, so the signal is difficult to achieve by ordinary color CCD. Here infrared CCD is employed as a probe, and obtained infrared image is input the computer, then fire recognition is determined based on image processing. Firstly, the grey scale binary image is partitioned by entropy threshold method and then smoothed by Median filtering. Then significant characteristics of flame are extracted and after normalization the foregoing data are input neutral network for recognition. All of the above clues are combined to reach a final decision. Experiment results show that the proposed method is successful in detecting fire or flames. In addition, it reduces the false alarms remarkably. It is also shown that the method works in a variety of conditions. The system has high validity and accuracy.

We demonstrate a diode-side-pumped continuous wave (CW) Nd:YAG laser at 1123 nm with over 100 W's output power and good beam quality. The resonator adopts convex-convex structure working in a thermally near unstable cavity. By precise coating design a single 1123 nm wavelength is delivered. Under the pumped power of 870 W, an output power of 106.5 W is obtained, corresponding to an optical-optical conversion efficiency of 12.2%. The beam quality of M(2) factor is measured to be 5.6. To the best of our knowledge, this is the highest output power with good beam quality for 1123 nm CW Nd:YAG laser.

Switched-mode DC-DC converters are nonlinear and time-varying systems. Chaos may occur in some case and it will influence the working performance even the entire system greatly. So chaos should be investigated in detail. A nonlinear discrete modeling technique is employed here. The system is described in terms of a sequence of sample , one per switching cycle, involving a mapping from one sampling instant to the next. The variation of the state vector within in a cycle is described by a function and it can be applied iteratively to determine how the state of the converter evolves from cycle to cycle. The proposed approach is verified by simulation and experiments, respectively.