Abstract: With noise in nature and blurred edge of image, edge detection and target segmentation is difficult. As a mathematical morphological method, soft morphological transform have many good characters of standard morphological transform and is robust to noise, so it is possible use it to analyze characters of target. In this paper, we construct edge detector with soft morphological transform and detect blurred edge in image. Experiments show, it can reserve detail of image and consequently remove noise, comparing with commonly edge detectors. It is very practical to many applications.