Search

With the rapid increase of on-site monitoring objects and technological development, the power on-site image monitoring data has increased geometrically. For security managers, it is necessary to efficiently and quickly identify valid data and conduct intelligent early warning. Therefore, this paper proposes an intelligent analysis method of power field surveillance video data based on convolutional neural network. Firstly, the original surveillance video material is extracted. Secondly, the video cloud storage platform is used for storage and data cleaning. At the same time, the hidden danger behaviors are classified and analyzed and marked. Finally, the proposed method is used for feature extraction, model training and field verification to realize intelligent early warning analysis of video big data. The field application effect shows that the method proposed in this paper can realize the functions of video image data feature extraction, model training, and improve the effect of image processing. Eventually, it can meet the needs of on-site video data intelligent analysis and has a certain promotion prospect.

Due to the complex site conditions of power infrastructure projects and the similar structural features of various danger sources, it is difficult to extract salient features suitable for classification. Although the neural networks with large numbers of available improved loss functions have good application effects in feature recognition, the single depth feature does not make full use of the complementarity between the multiple features of the field environment. In response to these problems, this paper proposes a method for classification and identification of danger sources at infrastructure construction sites based on neural network multi-layer feature information. It completes the image recognition of the infrastructure site by selecting the sparse self-encoding technology, extracts the multi-layer features in the neural network, and then sends the multi-layer features to the softmax classifier to solve the multi-classification problem of dangerous sources, and can extract more suitable based on the analysis of the characteristics of the danger sources on the infrastructure construction site. Through data analysis, the method illustrated in this paper effectively improves the identification rate of the danger sources and shows great significance to ensure the safety of infrastructure construction site.

With the development of energy Internet technology, the types and number of terminals in substations have shown an explosive growth trend, which puts forward higher requirements for image compression and image processing capabilities. Therefore, this paper proposes a method to improve the quality of power image transmission. Firstly, the eigenvalues and special vectors of the image are obtained, and then the wavelet transform method is used to encode the image, and finally the compressed image after decoding is obtained. Based on a typical field case, the comparison of experimental data under various algorithms is carried out. The results show that the improved method proposed in this paper has better image compression quality and higher peak signal-to-noise ratio, and improves the transmission quality of substation field monitoring images. It provides new technical ideas for the development of big data processing of electric power images.

To master the real-time security status of the infrastructure construction site, an intelligent monitoring system for the recognition of the danger sources at infrastructure construction site, which is based on the theory of depth learning, is proposed. The monitoring system mainly includes signal acquisition equipment, internal optical network, server control center and display terminal. The intelligent identification algorithm of the danger sources is the core of the entire system. The characteristics from the image signal are extracted by the sparse self-coding, which is used to train the neural network. Then, the convolution is proposed to reduce the dimensionality of the features. The experiments confirm that the identification algorithm based on deep learning illustrates high accuracy in hazard identification. In terms of feedback, the identification results will be transmitted to the display terminal, and the safety status of the entire infrastructure construction site can be fully controlled to ensure the safety of both the infrastructure site and the power system.

At present, massive amounts of image data have been generated at the power construction site. Due to the limitation of power communication transmission bandwidth, it is necessary to improve the low bit rate, high compression ratio, and anti-interference of the image, which is very necessary for on-site management personnel under the premise of ensuring efficient information transmission. Aiming at the above problems, this paper proposes an improved algorithm based on Set Partitioning in Hierarchical Trees (SPIHT). Firstly, the feature value of the image is extracted, and then the wavelet encoding is performed to obtain the compressed image. Compared with the traditional multi-level tree ensemble splitting algorithm, experimental data show that the image compressed by this algorithm has better visual effects and higher peak signal-to-noise ratio. Therefore, the method can effectively improve the transmission performance of the on-site image of the infrastructure construction, and ultimately improve the efficiency of the on-site inspection for violations.