This solution is divided into 4 layers.The bottom field instrument is the instrument already in the factory, which is used normally and needs to collect and display.
On the second floor, the camera is a dedicated time-capture camera, designed for harsh industrial environments. It can take pictures at regular intervals and upload the pictures to the designated FTP server. In this solution, the pictures will be uploaded to the factory's local intranetserver.
The third layer, network transmission equipment is the network transmission equipment such as wifi routers or switches deployed in the industrial field.
The top layer is the instrument identification software, which regularly reads instrument photos, performs OCR image identification, and stores the identified data in a designated database, such as MySQL SQLserver database.
About the introduction of recognition algorithm:
For the collected images, Gaussian Filtering is generally used for graphic noise reduction.For the preprocessed image, the method of Canny Edge Detection is used for edge detection. By accurately filtering various non-critical targets, the outline of the instrument and the panel can be found. The Hough transform is used to detect the scale of the instrument and the pointer of the instrument, and the liquid level can also be detected by the Hough transform.detection.For the targets in the image that need to be image classfition, deep learning methods are used, such as number recognition, instrument style recognition, and text position detection in natural scenes.The model adopts the deep learning medium convolutional neural network model CNN, and builds different network models according to the characteristics of the picture.
Pointer meter recognition is mainly for pointer meter, which is an automatic recognition of industrial meter recognition through video images..
The identification process is: adopt some special methods to complete the effective area screening and the pointer positioning of the instrument, and calculate the sub-dial according to the characteristics of the connection between the center of the dial and the center of rotation of the pointer and the angle between the 0 scale line of the sub-dial.The angle between the dial 0 scale line and the pointer pointing line segment is further identified and judged to read the pointer reading.The experimental results show that the positioning and recognition algorithm is simple in calculation, has a high accuracy, and overcomes the influence of the random angle of the dial on the reading recognition algorithm
本文为[iData Mining IoT-Shanghai/Henan]所创，转载请带上原文链接，感谢