The necessity of the learning type AOI automatic optical inspection systems is a problem that is worth exploring. Someone mentioned that the AOI equipment was capable of automatic learning, so it has a better effect for reducing the misjudgment, which is what we often refer to as the analysis method of AOI statistical modeling: correcting misjudged and misreported data to reduce false positives. All that needs to be done is manual confirmation. So, is the necessity of learning AOI really valuable? How significant is the necessity? This is the question that Xiao Hu will answer in this article.

The common user-friendly software design concept of the statistical modeling with common image comparison technology upgrades usually has no problem in the case of mass production, and the overall results are also good, so in the past few years, the development of AOI equipment has been popularized; however, the constant pursuit of progress has driven us to go even further on the road to PCBA solder joint inspection. Is it possible to save the efforts for learning or reduce false positives to near zero? This is the highest requirement for the AOI detection technology, as the current optical detection technology needs to be improved in the aspect of user friendly automatic modification of the element changes. For example: A newly recruited employee can plunge into production once being informed of the standards. Such production efficiency is the dream of every SMT manager. This is also the case for AOI equipment. If after the standard is given directly in the programming process of PCB, there will be no follow-up constant false positives and constant correction process, it is of course an ideal condition.

Contrary to the learning type AOI automatic optical inspection systems, the analytical type AOI is the abbreviation of the feature vector analysis type AOI often mentioned by Xiao Hu. Because the vector analysis type AOI is added with the function of automatic correction of the software in a user-friendly way, with respect to the subsequent problems such as positioning point deviations and FPC deformations, it can automatically correct the problem of high false positive rate of solder joints which may be encountered in image contrast technology based on the product features that we need to detect. The technology is described in detail in the chapter of “Vector Imaging – High-precision Positioning Technology, which will not repeated here. It is worth mentioning that the vector analysis technique fully utilizes different characteristics of different detection parts, and the feature is that the area occupied by the outline itself is an infinite number of pixels. The pixels are fixed and monotonous, but the features themselves are more abundant. At the same time, the vector analysis of the coordinate digital data is simpler compared to the image comparison and other picture digital data. The coordinate data is also more rigorous than the picture data in the analysis process, which determines that the new vector analysis technology bears inherent advantages. It has an excellent pedigree from the root of data analysis. If the image contrast is required to meet the requirements for improving low false alarm rate for PFC soft boards and plug-ins, it is necessary to start from the fundamental software analysis, which means that the underlying source software program should be changed. So far, no company has been found to specialize in the development of image contrast methods for optical inspection technology, or to rewrite its own procedures to meet the requirements of production inspection technology in the development of AOI detection technology.

In summary, the AOI of image contrast technology has played an invaluable role both at present and in the past. However, as AOI optical detection technology is increasingly developed, when production needs to question the process of the first-stage learning and debugging of AOI equipment, how much room for growth is there for learning AOI? Should our AOI detection technology be improved on the basis of image comparison, or should we overturn the old detection technology to acquire a newer knowledge on the new technology?