• Home
  • »
  • Blog Posts
  • »
  • How Camera Based Label Defect Detection Systems Work

How Camera Based Label Defect Detection Systems Work

Introduction:

In today’s food and beverage industry, packaging and labeling play an extremely important role in the presentation and subsequent selling time of products. When a product contains a label that is misprinted, skewed or in rough shape, it can lead to the product staying on the store shelf for much longer, ultimately leading to a longer sales cycle and money lost for the company. 

Additionally, labels play a major role in the trust and brand perception that consumers have regarding the products they use. Items with poor label quality are considered by consumers to be less trustworthy and ultimately can hurt a company’s reputation and brand image. 

A rigid quality control process ensures that labels printed on products are inspected for defects before they are sent off for sale. However, when this process relies entirely on manual inspection from employees it takes significantly longer and has a greater margin of error when compared to an automated vision based system. This is due to the fact that employees are prone to natural fatigue following hours of work and hundreds of items inspected, which increases the likelihood that they will miss a defect. A computer by contrast can work continuously to identify defects without requiring breaks or getting tired.

In this post, let’s take a look at how a camera based, defect detection system works to find issues in a label and how it takes the appropriate action to correct the problem. 

Central Components: The Hardware and Software

A vision based label defect detection system works by having two primary components – the camera for capturing images and the software which analyzes the incoming data. As the products move through the final stages of the processing line to quality assurance, each individual item enters briefly into the camera’s field of view for image capturing. 

At this stage, depending on the product, a strobe light may briefly illuminate the item allowing for the camera to capture more light, and produce a crisper image for processing. Once the camera, which is recording at 60 frames per second, captures these images they are sent off to the software where they are processed. 

Image Processing: Inside The Mind of The Operation 

The camera acts as the eyes of the operation, and the computer acts as the brain. The computer looks for specific defect criteria that are predefined and calibrated during the setup process in accordance with the product being inspected. As a focus point, let’s take a look at how the software would analyze and process an image of a barcode within the identified label to determine if it passed inspection, or failed. 

Depending on the calibration of the software, the processing algorithm first identifies an area defined as the location on the label which contains the barcode. In the viewfinder on the technician’s monitor, this area is identified as the bright colored square around the perimeter of the barcode. 

Following the identification of the barcode, the software first processes the image to improve features such as the brightness, contrast and sharpening to get a better read on the lines and white spaces which make up the barcode. This processing is usually done within the targeted area, as applying this processing to the entire image would increase the computing resources needed and subsequent time required for processing, ultimately slowing down the operation.

Next the image processing algorithm reads the barcode on the packaging and compares it to a set of predefined ones ‘trained’ into the system during the initialization process. Once it compares the two data points, the algorithm assigns a grading score to the image which can either be a simple pass or fail, a letter grade from A to F, a numerical scale or a combination of all 3.

For example, a letter grade scale may be from A to F, where anything rated A or B is considered a pass and anything rated C,D or F is considered a fail, and must be rejected. 

Depending on the rejection process put in place, the item may be automatically rejected and ejected from the conveyor line and put into a separate pile. Another rejection response, especially for more complex products such as technology components and pharmaceuticals is alerting the appropriate person (e.g a floor supervisor), to take appropriate action. This automated error flagging response is critical to ensuring that whatever the identified problem may be is resolved quickly, and with minimal product wastage. 

This entire process, from the product entering into the field of view of the camera, to being processed by the camera and assigned a grading score to determine if it is a pass or fails take place in under a second

The Results

When implemented properly, a label defect detection system can greatly increase the efficiency of an organizations’ quality assurance process. When compared to an employee, the camera and software do not fall victim to fatigue and can analyze items much faster. A properly trained system can also reach close to 100% in accurate detection rates, which is unparalleled in comparison. 

However it is important to note that if an automated system to reject the failed items is not in place, human intervention is still required. When an issue is detected by the system, a floor technician must use this information to take corrective action and resume the production line. 

Automated detection technology works best when there are clear data points to ‘train’ the software with. This may include reference barcodes, sample images of defects as well as photos of various item conditions. The software can also learn from identified false positives, and use what would normally be an error in the system as an additional data point using the more nuanced conditions that led to the false positive. The more information the software has to process the images, the more accurate its rate of defect detection.

Want to see how Alooki can help your organization implement a vision based label defect detection system?

Click the link to get in touch! 

#Big Data#Blog Posts

Share it!

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Posts

3 Benefits of An Automated Packaging Defect Detection System


Technological advancements in engineering and software capabilities have opened realms of new possibilities for applications of automated processes within virtually all  industries. One of these industries is the food processing and packaging industry where automation has been a cornerstone for decades. In food processing and packaging, some form of task automation is utilized in almost […]

Measure, Manage, and Create More Profitable Processes with Alooki – the next generation of making things smart. Leverage the power of Artificial Intelligence and Machine Learning for your organization!

Contact

1973 Topsail Road
Paradise NL
A1L 1Z4

Follow Us