Zhao Zhibing, Wu Di
Juicycode Team
With the rapid development of digitization and intelligence, intelligent recognition systems based on computer vision have been widely used in various fields. Among them, collection management and conservation is an area of great significance. The OpenCV-based Collection Treasure Intelligent Visual Recognition Management System (ArtIQ ™) aims to provide fast and accurate recognition and management of RMB and commemorative banknotes using computer vision technology.
ArtIQ ™ provides automated management and fast identification. While traditional manual identification of RMB and commemorative banknotes is time-consuming and laborious, the system utilizes OpenCV to implement an automated visual identification function, which can quickly and accurately identify the serial number and coin type, improving management efficiency. The system can record and store the recognized information, which can provide a large amount of data for the organization and facilitate statistics and analysis. Through the statistics of different coins and serial numbers, the organization can better manage and protect the precious collections of RMB and commemorative banknotes, etc., and improve the level and effectiveness of collection management.
ArtIQ ™ is developed in Python and the image processing functions are implemented using various functions from the OpenCV computer vision library. The software GUI is developed using PyQt5. The system is equipped with RMB and commemorative coin recognition function, which can quickly locate the RMB and obtain the parameters such as width, height, color, etc., intercept the set area and perform template recognition. "Color", "Template" and "Aspect ratio", which represent color recognition, template recognition and horizontal/vertical recognition respectively, are used by the system to identify RMB and commemorative coins. The system performs color, template and aspect recognition on the RMB image to get the final recognition conclusion in a weighted way. Based on ORB feature extraction, SVM is used to classify and recognize the commemorative coins.EAST is used to locate the serial number in the commemorative coins and then Tesseract is used to identify the serial number. There is complete access and storage management for the recognized image information.
The GUI is equipped with dynamic adjustment of weights, weight switch, output content control, dynamic adjustment of front and back recognition binarization parameters, dynamic adjustment of template recognition sensitivity, and matching rule switch, and at the same time, it is equipped with the function of embedding the window to display the "Final Recognition View", the "Original View before Correction", and the "Binarization Real-time Processing View" view, and can display one of the above three views by switching. At the same time, it has the function of displaying "Final Recognition View", "Original View before Correction" and "Binarization Real-time Processing View" embedded in the window, and it is possible to display one of the above three views by switching. The "Open All Views" toggle option is also available, allowing the user to open all process views, including image cropping. In addition, there is a real-time work output window that prints recognition results and system prompts in real time.
The system also has a real-time warning system that can identify potential errors based on the user's parameter modification behavior. When the user's adjustment of weights results in the weight of one weight being greater than the sum of the weights of the other two weights, a prompt window will pop up, and the user only needs to click on the "√" button to restore the parameter to the recommended parameter and solve the problem. You can also close the prompt by clicking the "X" button and choose to ignore the prompt until the next startup of the recognition.
In order to control the stability of the light source and background during the recognition, we have customized the "Juicycode Environment Control Light Box", which is equipped with three light source modes, namely yellow light, pure white light and yellow-white mixed light, and supports 12-step gradient light intensity adjustment. The light box is equipped with a light-absorbing cloth and a camera slot, which provides the most stable recognition environment and improves the accuracy of the recognition. The light box is equipped with light-absorbing cloth and camera slot, which can provide the most stable recognition environment and improve the recognition accuracy.
