Beihai Tan

Bio: Beihai Tan is an academic researcher from Guangdong University of Technology. The author has contributed to research in topics: Character (mathematics) & Stand-alone power system. The author has an hindex of 3, co-authored 6 publications receiving 22 citations.

Method and device for recommending online answer question

Abstract: The present invention discloses a method for recommending an online answer question. The method comprises the following steps: receiving an online answer request of a user for a target question library; obtaining historical behavior data of the user; determining a to-be-output question according the historical behavior data; further determining a question output order, and in the question output order, outputting to-be-output questions sequentially; during an online answering process of the user, determining whether a current question meets a preset associated question recommendation condition; if yes, determining an associated question of the current question in the target question library; and recommending the association question to the user. According to the technical scheme provided by the embodiments of the present invention, the to-be-output question that is determined according to the historical behavior data better meets the answering intention of the user, and during the answering process, the associated question is recommended to the user, so that the user can perform knowledge-point consolidation excercises more pertinently, and the answering experience of the user is improved. The present invention further dislcoses a device for recommending an online answer question, which has a corresponding technical effect.

Character recognition based on corner detection and convolution neural network

Abstract: Text detection and character recognition plays a very important role in the field of computer vision. Although there are many studies of character recognition, the existing text detection methods are mainly concentrated in the English characters, and now there is a great application needs in Chinese text detection. As Chinese characters are more complex than English, Chinese character detection requires more efficient technology. Aiming at the English and Chinese characters in the picture, this paper presents a character recognition method based on corner detection and convolution neural network. Firstly, we use image processing technology to preprocess the input image, and then use the corner detection method to mark the text candidate area. Secondly, the image histogram of oriented gridients (HOG) feature extraction and support vector machine technology are used to filter the text candidate area. Thirdly, the integral projection method is applied to split the character area as single character blocks. Finally, the model of convolution neural network is used to identify the segmented character blocks. The experimental results have shown the effectiveness of the proposed method in the test image set.

Recommend Algorithm Combined User-user Neighborhood Approach with Latent Factor Model

Abstract: The item-item neighborhood model became very volatile for current items rapidly replaced, such as online article and news items. And the neighborhood model faces the problem of data sparsity and cold start. The factor model can alleviate data sparseness problem, but it does not take the historical behavior data into consideration. Therefore, in this paper, we proposes a recommendation algorithm based on user-user neighborhood model and latent factor model, which can make accuracy improved significantly and can effectively address the problem of data sparsity. When the number of neighborhoods k increases, the accuracy of the algorithm has improved. The experiment result shows that this approach is correct and feasible.

Islanding detection for PV plant using instantaneous power theory

Abstract: Islanding detection is a distributed generation system is an important protective measure. Islanding detection is generally divided into two types, one is the active islanding detection, and the other is passive islanding detection. Islanding detection of power quality disturbance signal will always cause pollution; passive islanding detection has a long detection time and a large dead zone. Various existing detection methods have blind spots, resulting in test failure. In this paper, islanding detection blind, it raised the wavelet coefficients energy distribution and energy fluctuation coefficient analysis. Then combined with passive islanding detection, reduce or even eliminate blind spots. In this paper, MATLAB simulation and verification method proposed introduction of a device. The simulation results show that verification, islanding detection algorithm proposed in this paper has a good real-time performance, reliability and stability. The outcome can be applied to a large number of islands to detect the actual PV power plant, or micro-grid islanding detection system.

An attention-based network for serial number recognition on banknotes

Abstract: The serial number recognition (SNR) on banknotes is essential for currency circulation. The performance of the existing SNR methods is significantly influenced by character segmentation, which is challenging due to uneven illumination and complex background. In this paper, we apply deep learning techniques to SNR by proposing an attention-based network, which can be end-to-end trained to avoid the problem of character segmentation. The proposed framework contains two parts: rectification and recognition. First, the rectification network, which can be trained in a weakly supervised manner without additional manual annotations, is built to automatically rectify the tilted and loosely-bounded images and reduce the difficulty of recognition. Then, the recognition network, an attention-based sequence model, recognizes serial number characters from the rectified image holistically, rather than segmenting and recognizing individual characters. To address the problem of complex textures on banknotes, we integrate the deformable convolution into the recognition network, which adaptively focuses on the character regions by using flexible receptive fields to accurately extract optimal character features, while ignoring redundant background information. Extensive experiments conducted on CNY, KRW, EUR and JPY banknotes, demonstrate that the proposed method achieves higher accuracy than the existing methods.

Performance Evaluation of Maximum Power Point Tracking Approaches and Photovoltaic Systems

Abstract: This paper elaborates a comprehensive overview of a photovoltaic (PV) system model, and compares the attributes of various conventional and improved incremental conductance algorithms, perturbation and observation techniques, and other maximum power point tracking (MPPT) algorithms in normal and partial shading conditions Performance evaluation techniques are discussed on the basis of the dynamic parameters of the PV system Following a discussion of the MPPT algorithms in each category, a table is drawn to summarize their key specifications In the performance evaluation section, the appropriate PV module technologies, atmospheric effects on PV panels, design complexity, and number of sensors and internal parameters of the PV system are outlined In the last phase, a comparative table presents performance-evaluating parameters of MPPT design criterion This paper is organized in such a way that future researchers and engineers can select an appropriate MPPT scheme without complication

Power & energy optimization in solar photovoltaic and concentrated solar power systems

Abstract: Solar Energy purvey a pure Environment-friendly, ample and everlasting energy resource to humanity. Electricity can be generated using solar energy by two different technologies namely photovoltaic (PV) and concentrated solar power (CSP) systems. By using thermal energy storage technologies, CSP systems can store energy to generate electric power on cloudy days or overnight as compared to PV systems which results in flexibility in power network. Most important issue in energy market is the competitive cost of energy. Energy price of PV plant is less as compared to CSP plants. Whereas, CSP systems with thermal Energy storage capabilities can be effectively used to overcome intermittency issues of PV systems to balance demand with the supply of Electric power within safe levels of reliability by optimizing the Energy produced. This paper try to figure out the possible ways to optimize power and energy produced by Solar Energy technologies to reduce Carbon footprint. In addition to that, Solar PV and CSP systems are compared at two locations of tropical country India and simulations has been done in System Advisor Model (SAM) and presented graphically.

An examination question recommendation method and tutoring device

Abstract: The invention relates to the technical field of electronic equipment, in particular to an examination question recommendation method and a tutoring equipment, comprising the following steps: the tutoring device obtains the answering time length of a user for a target topic, and judges whether the answering time length is greater than a preset time length threshold value; if the answering time is longer than the preset time threshold, the exercise questions whose similarity with the target questions is greater than the preset similarity threshold are obtained, and the exercise questions are recommended to the user. By implementing the embodiment of the invention, the learning effect of the user can be improved.

Segmentation and identification of presence of Sinhala characters in Facebook Images

Abstract: With the rapid growths of social network, people are connected to one platform and build closed relationships without caring any geographical barriers. Facebook, Twitter, Google+, LinkedIn can be seen as most popular social media. By studying the content on social media, can get views of the society of a current topic which is talked in the society. In Sri Lanka, there are large number of Sinhala typed optical image posts are shared by public groups to the Facebook. To study those content, first it is necessary to identify Sinhala content images. All images are not containing characters and all characters are not Sinhala. So it is necessary to filter only images with Sinhala characters. One way of identify the availability of Sinhala characters is by doing the whole process of Object Character Recognition. But when a large number of images are processing, it is a disadvantage to have non-character and non-Sinhala character images with the Sinhala content images. So it is better to filter only Sinhala character images to do the process. But there were no researches to identify the availability of Sinhala character is an image. So this research was done to identify the availability of Sinhala characters in an image. Facebook posts which were published by the public groups are taken as the domain to the research. To identify the availability of Sinhala characters in an image a Rule-Based method was proposed in this research.