Fabrics are susceptible to defects caused by various factors during the dyeing and finishing processes. Fabric defects are mostly detected through manual visual inspection. Under the influence of fatigue from long working hours and human error in general, this method results in misjudgment and missed defects. Certain difficulties are encountered in the development of automated optical inspection (AOI) machines because large pieces of fabric must be inspected at high speed. In this study, an AOI system was developed. The image acquisition system consisted of an IDS industrial camera equipped with a sensor for fabric defect inspection and a line scan function, as well as a white light-emitting diode (LED) light bar. The parameters were configured and calibrated using uEye Cockpit software. The inspection mechanism of the proposed machine was a roller device that simulated fabric production. Under continuous rotation of the roller, images were captured using the camera and then were combined to form a two-dimensional plane image. Long-term inspection was performed simultaneously. The line scan settings were a pixel clock (PCLK)S rate of 474 MHz and 10 scan lines per single scan when the tangential rotation velocity of the roller simulation was 0.9 m/s. The minimum vertical distortion of the acquired image was 1.658 mm when the total image length was 210 mm. Uneven illumination from the nonlinear light source resulted in the generation of noise in the captured images, which was eliminated through the combination of posterization and closing, with only a slight loss of defect features. Overall, the proposed AOI system effectively detected the defects. The minimum width of black defects that could be detected in the dyed and finished dark fabrics was 0.24 mm.