The invention belongs to the technical field of coal seam macroscopic fracture prediction, and particularly relates to a coal seam macroscopic fracture dominant development orientation prediction method, which comprises the steps of 1, selecting a coal wall picture, processing an original coal wall image by adopting different gray thresholds, taking white as an area with a gray value smaller thanq, and taking black as other parts; 2, carrying out information input and output, including single fracture picture input, multiple fracture picture input and fracture information output; 3, preprocessing the image, wherein as the underground shooting condition is limited, crack pictures often have poor definition and are reasonable in structure, and noise reduction processing, binarization optimization, image edge detection and the like can be automatically carried out on the preprocessed pictures, so that the digital pictures reach the actual application standard, the tedious workload of workers can be greatly reduced, and the efficiency and accuracy of field recognition are improved.