Improving the quality of data recorded by the machine vision system of a contact welding machine for semiconductor crystals

Modern advances in electronics make it possible to create data processing units with high energy efficiency and compact dimensions. The need to sense and develop units capable of analyzing external space has become an essential element of modern devices. The development of adaptive and self-organizing systems is impossible without data processing and analysis devices. The use of sensor units as an element of spatial communication and the formation of their mapping into the domain of information parameters is associated with their transformation and simplification. This interaction introduces noise into the data, the control of which, even in modern technology, remains an important and pressing task. Aim. Development of adaptive approaches to data processing using multi-criteria methods and the formation of implementations to improve the accuracy of machine vision systems (MVS). Materials and methods. A multicriteria processing method based on the minimization of a combined criterion is presented. This method enables the implementation of edge detection, noise smoothing, and background/object region extraction. An implementation for processing two-dimensional signals in localized regions is proposed. Results. A flowchart of a data processing algorithm based on a multicriteria objective function is proposed. This algorithm enables processing of both images and one-dimensional data arrays. The ability to locally process images containing objects with sharp transition boundaries improves efficiency compared to the standard implementation. Data MVS of a semiconductor crystal bonding machine prototype are used as test data. The proposed algorithm improves the distinguishability of element structures, reduces noise, increases mask construction accuracy, and enhances the visual quality of the data. Conclusion. The proposed solution improves the quality of vision data for a semiconductor component chip bonding system layout. This enables smoothing of noise components in locally stationary image areas, while preserving transition boundaries, markings, and special symbols.