The invention provides a machine vision-based elevator compensation chain place-R crack detection method. The method includes the following steps: (1) collecting an image which is a chain place-R crack image, and carrying out partitioning on the image, wherein image partitioning comprises threshold segmentation for extracting a centerline of an elevator compensation chain and a place-R area, whichcontains to-be-detected defects, to obtain a to-be-detected target area image of the place-R area; (2) carrying out corresponding local enhancement, filtering and noise removal and edge extraction onthe to-be-detected target area image, which is collected in the step (1), of a place R, and then carrying out area enhancement processing on the place R to obtain an ROI area image; (3) carrying outthreshold segmentation, open operation and morphological feature extraction on the ROI area image obtained in the step (2), primarily determining the defects thereof, and extracting candidate defect images; and (4) carrying out secondary determination on the candidate defect images, which are obtained in the step (3), according to location features to obtain a final defect image determination result. According to the method, an automation degree is higher, detection is accurate, and detection efficiency is high.