Visual sensing system + parallel robot: fast identification and sorting PCB board

It is well known that the 3C industry belongs to the "fast pace" industry, and its "fast" is mainly manifested in 1. The continuous launch of new products, so that the product update cycle continues to shorten. 2. Technological innovation is applied in the field of 3C, 3C products are constantly developing towards the intelligent direction, and downstream development promotes the continuous upgrading of relevant production equipment.

As the demographic dividend disappears and labor costs continue to rise, the automation of the 3C industry is an inevitable trend. Industrial robots, as an important part of 3C industry automation, undertake the task of "machine replacement". Among them, parallel robots have been characterized by fast speed and high accuracy, which can be used for sorting, sorting, line transfer, loading, and other technological processes.

Taking the PCB board as an example, the PCB board is of various types, with different shapes and similar colors, which is difficult to be distinguished by the naked eye. Moreover, the layout and wiring of the printed circuit are miniaturized, high density, and line spacing. The traditional manual feature description has limited ability, and the features designated by humans are usually less in number, single in information, poor generalization ability of description, and susceptible to noise interference. This process requires not only high reliability, high precision, and high speed of the parallel robot, but also strict vision requirements of the robot.

To this end, the vision sensing system independently developed by Birkent, a member of the G20 Intelligent Manufacturing Summit, is equipped with parallel robots, which can quickly identify and sort hundreds of PCB boards. The visual detection system based on deep learning can automatically detect PCB through camera shooting equipment.

When the circuit board is made from the plate-making equipment, after the completion of the corresponding test equipment for PCB testing, to determine whether products or defective products, and then judge the results back to the parallel robot vision system, when the conveyor belt will be sent to PCB vision camera, visual, the opportunity to identify non-conforming signals to parallel robot to eliminate sorting, identification of qualified products The signal is transmitted to the robot for good product sorting and sorting, and passed on to the next process.

Based on deep learning, the Birkent visual sensing system can automatically learn complex feature descriptions from large-scale data sets. Deep learning overcomes the blindness of manual feature extraction and has a strong anti-noise ability. Suitable for high-precision parallel robot belt tracking scenarios.

When the target is in the camera view, visual software will receive read external camera trigger signal sent by a robot, phase target image real-time acquisition, through a particular image processing algorithm, complete the target contour extraction, target orientation, shape recognition, and color recognition, and other functions, and these objects characteristic information transmitted to the robot, the robot according to the motion control algorithm, to complete Tracking and grabbing function for materials.