Abstract: Aiming at the problems that the traditional protection platen image recognition method has strict requirements for lighting conditions and low recognition rate, and the current hot deep learning method for identifying the protection platen state has high requirements for computer hardware and large sample size. An improved method of color extraction and feature recognition of platen state is proposed. The screen and cabinet diagram collected by the mobile end of the inspection robot is preprocessed to improve the clarity and recognition of the image. Then the color image is transformed into the YC_bC_r color space for color feature extraction, the binary image is obtained, the binary image is morphologically processed and the connected domain is extracted. Then, the morphological characteristics of the connected domain are analyzed, the signboard and small interference spot are removed by using the area and boundary information analysis, the adhesive connected domain is identified by using the limit corrosion algorithm, and the adhesive connected domain is segmented. Finally, the simple minimum circumscribed rectangle and the area minimum circumscribed rectangle algorithm are used to obtain the circumscribed rectangle of the connected domain, locate the pressing plate, and judge the state of the pressing plate according to the length height ratio and deflection angle of the circumscribed rectangle. The test results show the accuracy of the proposed method for pressing plate cutting, the generality of recognition for special situations in complex environment, and the efficiency and accuracy of the recognition method.