Recognition Of Insulator Crack Types Based On Neural Network And The Invariant Moment

The intelligent patrol robot of High-voltage transmission line inspection produced with the transmission engineering and technology development. The robot systems can observe the power lines and transmit the video signal to supervision station automatically. The style-working will greatly ease people's work condition and improve the patrol efficiency and reliability. The development of digital image processing and artificial intelligence makes it possible for the High-voltage transmission line intelligent inspection. The studied title is come from national "863" fund projects: Intelligent Controller for High-voltage Transmission Line Charged Robot.Based on image recognition, insulator cracks and achieves the classifications of the cracks can be detected by simulating vision system of robot. The images information can be captured by the hardware, and then completes the image pretreatment, the image segment, the feature extraction and the classified recognition.To the image segment, a method is put forward that the image can be divided by pixels of sub-block image in this paper. In addition, this method can also distinguish crack and non-crack goal. The global features of insulator cracks image can be extracted by the invariant moments of sub-block image matrix. Last, the classifications are designed based on BP neural network, which identify insulator cracks of three classifications: Single cracks, massive cracks and mesh cracks.