Steel wire rope reel irregular arrangement detection method based on computer vision

The invention provides a method for detecting irregular arrangement of a steel wire rope reel based on computer vision. The method comprises the following steps: step 1, acquiring a real-time video of the steel wire rope reel on a construction site of a tower crane; 2, extracting image data for preprocessing; 3, inputting the preprocessed image into a steel wire rope reel anomaly detection model; 4, judging whether the current steel wire rope reel has rope disorder abnormity or not, and if yes, carrying out voice prompt; the step 3 comprises the following steps: S3.1, inputting the preprocessed image into a steel wire rope drum anomaly detection model, and optimizing a FasterRCNN algorithm by the model based on a deep learning thought; s3.2, carrying out optimization and improvement on the original image position coordinates of the FasterRCNN algorithm based on an ROI-Align algorithm; s3.3, designing an improved ROI pooling layer algorithm; s3.4, feature extraction of the convolutional neural network is optim