Hyperspectral/Multispectral Imaging Methods for Quality Control

Introduction

According to Crosby, the quality of a product is defined as its conformity to the specification (Crosby, 1979). There are various processes employed in production environments to assure product compliance to its specifications. Visual product inspection is one of such methods used in production environments to assure the absence of defects in products. Mostly, the product quality inspection is done by human operators visually on factory floors. However, human inspection is unreliable as there are various factors affecting product quality determination. Different human operators may have different judgements in their product quality inspection due to work experience, conditions at the work environment, level of understanding of the product, psychological factors, fatigue, biological factors of the worker. Products with more than one type of defect take a longer time for visual inspection. When the acceptance criterion of defects changes, it complicates visual inspection. The human inspection has more downfalls with qualitative measurements as results for the qualitative measurements are varying and difficult to compare. Moreover, human perception can easily accept false positives (Kerkeni et al., 2016). As a result, human inspection is inconsistent, subjective, and slow.

FurthermoreMoreover, the visual inspection is incapable of determining physicochemical characteristics of the product, such as moisture content, presence of various microorganisms, texture, etc. The physicochemical characteristics estimations are vital to ensure the quality of different products such as pharmaceuticals, food, and beverages. There are various analytical methods used in production floors to estimate those physicochemical parameters of products.

The process control methods can be distinguished into four categories: in-line, on-line, at-line, and off-line methods. In-line methods are directly immersed into the process flow, while on-line methods use a bypass channel from the main process flow. The at-line methods analyse the samples next to the process flow by sample extraction and off-line methods process analysis separately from the process flow by withdrawing some samples (Boldrini et al., 2012). Even though some manufacturing processes can employ off-line quality evaluation methods by taking random samples, they are not suitable for quality critical production processes. The food and beverages manufacturing, pharmaceuticals manufacturing industries require a complete quality inspection prior to human consumption, requiring in-line quality checks. Again, in some cases, random sample testing is not suitable as it may destroy the product. In such cases, non-destructive in-line process control is a vital requirement in ensuring product quality.

Therefore, it is proved that the industry needs consistent, accurate, fast, cost-effective and in-line automated inspection methods to increase the effectiveness of visual inspection. With the recent developments in imaging and computer technology, machine vision solves most quality inspection and control needs in production environments.