The invention provides a knife switch state refined detection and identification method comprising the following steps: inputting a knife switch image to be detected, and preprocessing the image; inputting the image into a pre-trained deep learning module to judge the opening and closing states of the knife switch, and if the knife switch is in the opening state, directly outputting a judgment result and ending the detection; if the knife switch is in the closed state, judging the type of the knife switch to be detected, if the knife switch is a double-arm folding knife switch, sharpening theimage, extracting a characteristic value in the image, carrying out characteristic coding on the characteristic value, then identifying to obtain position information of two vertical double arms according to the characteristic code, calculating the closing degree of the knife switch, outputting and ending the detection; and if the knife switch is a double-column rotary knife switch, carrying out edge detection on the image, selecting a circular area on the edge to obtain rough position information of the knife switch, carrying out linear detection and positioning on the image to obtain accurate positions of two sections of conductive knife switches, calculating the closing degree of the knife switch, outputting and ending the detection.