High Precision Medicine Bottles Vision Online Inspection System and Classification Based on Multi-Features and Ensemble Learning via Independence Test
To address the problem of online automatic inspection of drug liquid bottles in production line, an implantable visual inspection system is designed and the ensemble learning algorithm for detection is proposed based on multi-features fusion. A tunnel structure is designed for visual inspection system, which allows the bottles inspection to be automated without changing original processes and devices. A high precision method is proposed for vision detection of drug liquid bottles. Local background difference method is utilized as a soft on-off to capture bottle image. And an image gray level equalization pre-processing technology is used to eliminate the impact from illumination. Three features are designed, which contain blocked histogram of gradient, blocked histogram of gray and Raw-pixel. An ensemble learning algorithm is proposed based on independence test and multi-features fusion, after theoretically analyzing the precondition of precision boosting to ensemble learning. Some results of analysis and comparison prove that the proposed method is advanced compare with baseline methods. And specially, there exist remarkable advantages in our method, when there are some noises in sample labels. Then, we carried on a 72 hours continuous test on a practical production line, in which the error rate of inspection is less than 1‰. In terms of time and precision, it is superior to the traditional manual detection. And hence the test results prove that the visual inspection system designed and algorithm proposed are advanced and practical.
This dataset is used to train the algorithm