Study On Numeric Recognition Technology And Movement Object Detection In Substation

The remote video surveillance system is an important part of the substation integrated automation system. Because the system can'see'the situation with cameras, it can help even take charge of officers on duty to comprehend the situation of the appearance and operation of electrical equipments and plants in the substation, the station environment, and so on. The remote video surveillance system added the smart analysis function can automatically deal with the relevant information. The system becomes more intelligent, and improves the degree of substation automation, as well as truly provides conditions for the realization of the unmanned substation.The numerical identification of digital instrumentations and objects movement detection in substation are main contents of the article. The paper introduces the specific applications of image recognition technologies in the substation and illustrates the digital identification process. The process of an image got by the camera is turning the RGB image into gray, positioning, binary process, filtering, correcting, single figure segmentation based on improved projection and normalization. Because the region of a digital image may contain a decimal point and studies of the point message is little, the article presents a new method of dealing with the decimal point. Digital images and the decimal point position information are sent into BP (back propagation) neural network to identify. The second identification is carried out if the first fails. According to recognition results, whether need to carry on alarm processing or not is determined. Finally, the digital recognition algorithm is evaluated. The experimental results show that image processing and recognition method the article used have high accuracy rate and reliability, and the method of dealing with the decimal point is feasible.Substation environmental monitoring is one of main functions for the remote monitoring system and is based on the movement detection algorithm. Current movement detection algorithms commonly used are compared. Combined with the optical knowledge, a threshold segmentation method based on cumulative histogram of the ratio image is proposed to detect the anomaly information of the image. Moving targets are tracked if the image detects the movement. The initial background got by restoration and two reference frames are adapted to update and support the background model. And the motion detection algorithm is evaluated. Finally a new concept of coordinated monitoring in the orthogonal direction is put forward. Experiments prove that segmentation results of the threshold segmentation method proposed in the paper are relatively good, and can get relatively complete information, and the anomaly detection rate of the object penetrated is high. Experimental results also show that the video analysis algorithm can adapt to the gradual light and ray mutation, and reduce the false alarm caused by some changes of the background (such as changes of the door) as well as light caused by mutation.