MULTI-CHANNEL CNN BASED METHOD FOR DETECTING POWER TRANSFORMATION FIELD OPERATION VIOLATIONS

The present invention provides a multi-channel CNN based method for detecting power transformation field operation violations. The detection method comprises the following steps: first collecting power transformation field operation violation images, calibrating violation categories to which content of images belongs, and constructing a data set of violations; then training a multi-channel CNN model on the basis of the data set, the multi-channel CNN model being composed of multiple channels, each channel using convolution kernels of different scales to extract image features, and improving detection capability for targets of different scales by fusing multi-channel information; and finally, preprocessing the operation images collected on a power transformation field, and inputting the preprocessed images into the multi-channel CNN model to implement violation detection based on the multi-channel CNN. The multi-channel CNN based method for detecting power transformation field operation violations provided by