Conventional edge detection technology can recognize boundaries of faults，fractures and sand bodies，but it is difficult to meet the demands of description of the boundary response of the geological bodies in different scales. To this end，an integrated multi-scale edge detection technology is proposed to precisely identify small faults and fracture zones，which combines three-parameter wavelet transform with a structure-oriented gradient operator，with dip volume and azimuth volume as constraints. The faulted-block reservoirs and Carboniferous weathering crust fractured reservoirs are dominant in Block Well P691 of Chepaizi area，western margin of Junggar Basin. However，limited by the poor seismic data，conventional methods cannot accurately identify faults and predict fracture development. Application results demonstrate that the designed technology can recognize small faults and predict fracture zones on the basis of accurate location of edges and multi-scale detection. Large and small faults and fracture zones can be clearly identified on the large-scale，mesoscale and small-scale edge detection maps，proving the effectiveness of this technique.