A set of tool wear monitoring methods based on machine vision was proposed aiming at the wear form of twist drill, in order to monitor the wear conditon of the tool during the machining process. A non-local mean denoising method was proposed based on integral image and Turky bi-weight kernel function according to the gray distribution of worn tool images. The single and double threshold Otsu methods were proposed to obtain the gray interval of the worn area to adaptively enhance the image. A method of extracting local extreme points of wear regions based on morphological reconstruction was proposed to effectively complete the detection of wear regions and boundary extraction. Experimental results show that the monitoring method of tool wear can be effectively implemented in the monitoring of twist drill wear. And it is proved that the proposed method has higher detection accuracy and efficiency and it can extract tool wear more accurately than other methods. It helps to realize the monitoring of tool wear and automatic monitoring of the processing process, and achieve the purpose of reducing labor costs and product failure rate.