Defect Detection of SLM Parts Based on Adaptive Threshold Segmentation Method

Parts manufactured by selective laser melting (SLM) technology usually have high surface roughness, porosity and residual stress, which seriously affect the mechanical properties and fatigue performance of parts. Therefore, efficient detection technology plays a crucial role in the quality and process of SLM parts. This paper proposes a threshold segmentation method based on object variance weighting, which solves the problem that the traditional global threshold segmentation algorithm is prone to produce wrong threshold results when processing unimodal images. Based on the traditional Otsu method, this method introduces an adaptive weight coefficient to make the threshold adaptively applicable to different unimodal histogram distribution images. By introducing this weight coefficient, it can be ensured that the image threshold is always on the lower left side of the peak in the case of a unimodal histogram. To verify the effectiveness of the algorithm, this paper analyzes the image by calculating the misclassification error (ME) and intersection of union (IoU). The experimental results show that compared with the Otsu, VE, NVE and WOV segmentation methods, the ME value of the proposed algorithm is closer to 0, and the IoU value is closer to 1, showing higher accuracy and better segmentation effect.