Objective: To solve the problems caused by the manual sampling inspection in the field of food packaging, such as difficult to operate continuously for a long time, easy to miss and wrong detection, and unreliable detection accuracy and stability. Methods: In this paper, a machine vision-based vacuum sealed texture detecting method for transparent packaging bag was proposed to replace manual detection. The image was preprocessed by algorithms such as ROI extraction, affine transformation and local binary pattern to highlight the texture features. On this basis, the gray level co-occurrence matrix was used to analyze the features of "good" and "defective" sealing texture images. The parameters of gray level co-occurrence matrix were set and the uniformity of texture features was associated with the feature quantity of the parameters of gray level co-occurrence matrix. Finally, the parameters of gray level co-occurrence matrix was used as the input of SVM classifier, and the sealing defects were identified and classified through calculation. Results: This online detection method compares the defect detection results of the vacuum sealing of transparent packaging bags with the manual quality results up to 97.5%. Conclusion: This method has high detection accuracy and good practicability, and can meet the needs of online detection.