Traditional Chinese medicine residue sorting and quality grading method based on machine vision recognition
By establishing a joint model of camera and light source response and performing state space scanning using an improved VMamba model, combined with the adaptive update of the TENT algorithm, the inconsistency problem in the sorting and quality grading of Chinese medicinal residues was solved, and stable sorting and grading were achieved under the condition of light source and camera drift.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEBEI UNIV OF TECH
- Filing Date
- 2026-04-22
- Publication Date
- 2026-05-26
AI Technical Summary
Existing technologies struggle to maintain the stability and consistency of sorting and quality grading of medicinal herb residues without interrupting production, especially under conditions of light source aging and camera response drift, leading to inconsistent sorting and grading results.
By collecting a continuous sequence of images of Chinese medicinal residue on the conveying line, recording exposure, white balance gain, and light source driving quantity, a joint response function model of camera response and light source response is established, a frame-by-frame system transfer function sequence is generated, and transfer function inversion compensation and reflectivity uniformity are performed. Combined with the improved VMamba model, asymmetric state space scanning and cross-scale state transfer are carried out. The TENT algorithm is used for label-free adaptive updates, the parameter update amplitude is limited, and an adaptive normalized parameter set is generated for sorting and grading.
It improved the consistency of imaging photometric properties and batch-to-batch comparability, significantly enhanced the consistency and robustness of characterization of Chinese medicinal residues, strengthened the consistency and reliability of sorting category determination and quality grade determination, and improved the reliability of diversion control command generation and graded recording data.
Smart Images

Figure CN122090426A_ABST