Research progress on key technologies of pharmaceutical intelligent manufacturing production line

In the development process of visual robot for medical quality inspection, the existing visual foreign body detection methods are difficult to meet the requirements of online detection of weak medical foreign bodies with various types, diverse features and high speed and accuracy. Therefore, the mechanical structure design of the medical quality visual inspection robot, image acquisition of weak foreign body targets under complex background, visual detection and recognition and other technologies are studied. It is the key and difficult point of medical quality visual inspection robot.

2.2.1 Mechanical structure

The mechanical structure of pharmaceutical quality visual inspection robot is directly related to the performance of drug image acquisition and detection methods, and its scientificity and rationality are particularly important.

At the same time, the moving parts can be integrated and analyzed in the digital environment to solve the kinematic constraint equation between the moving mechanism, and the simulation operation, process optimization, spatial interference analysis, collision analysis and other experiments can be carried out on the integrated digital platform, and finally the structure can be improved and optimized according to the experimental results. The whole mechanism of the robot can be analyzed from the aspects of design, processing, material, assembly and so on, with high precision, high symmetry, dynamic balance and global stability during operation. Then, according to the changeable packaging forms and specifications of drugs, the body structure of multi-type pharmaceutical foreign body visual detection robot is designed, which can realize the detection of different specifications and models of pharmaceutical products.

Service coordination control system

The drug image acquisition system not only directly affects the quality of the image, but also indirectly affects the performance of the detection algorithm. Therefore, for the weak foreign object image acquisition scheme under complex background, a series of repetitive action commands such as "precise alignment - flexible bottle grab - high-speed rotation - emergency stop - image acquisition - foreign object detection - sorting" are usually executed by the manipulator, and then the image is acquired by the CCD camera.

In this process, due to the existence of mechanical vibration, the camera will be inaccurate in focusing and the obtained medical detection image will be blurred, which will increase the false alarm rate of detection. In addition, with the increasing production demand, the pharmaceutical industry has put forward a higher demand for the speed of medical quality detection, and a single CCD camera is difficult to meet this requirement. In the multi-manipulator visual servo cooperative control system, by setting several high-resolution image sensors on the single tracking swing arm and optimizing the control time sequence of rotation-stop-tracking shooting, both high-resolution and stable images can be obtained and the detection speed can be improved. By optimizing the variable control parameters such as bottle clamping, bottle rolling and speed, the manipulator can simulate the effect of human bottle rolling in a fast and smooth mode, and improve the stability of bottle rotation.

Through the use of distributed network intelligent control system of vision inspection robot, including multi-channel vision sensor image acquisition and processing, multi-channel robot bottle clamping and bottle rolling optimization control, product classification and packaging control, etc., the opto-electromechanical system can ensure the coordinated work.

2.2.3 Visual detection and recognition algorithm

In the complex pharmaceutical production process, the occurrence of drug quality defects such as foreign bodies and packaging damage is unavoidable. Foreign bodies will directly affect the therapeutic effect of drugs, and even threaten the health of patients. Packaging damage can also directly damage the sterile environment of the drug and indirectly affect the quality of the drug. Therefore, visual detection and recognition technology is very important in the medical quality inspection.

Foreign bodies usually refer to some weak foreign bodies that are not obvious, such as silk thread, transparent plastic, glass fragments, etc., whose images have the characteristics of low contrast and low energy. Since foreign objects are small in size and most of them are transparent, the grayscale of the actual image does not change much relative to surrounding pixels, and the contour is fuzzy. In addition, external noise interference also makes the entire foreign object detection process more difficult. Therefore, in complex environments, research on detection and recognition methods with strong anti-interference and high robustness is the key to solve the problem of tiny target detection. It can greatly reduce the system error rate. Traditional methods usually filter, segment, edge extraction and other operations on the original image of foreign bodies, then select appropriate feature parameters, and finally classify and recognize foreign bodies through classification algorithms.