OCR Verification and Character Inspection

Parts and products are marked with expiration dates or product numbers. However, this is not useful when this information cannot be read, recognized, judged, stored, and managed accurately. Character inspection has become widespread in recent years for recognition or judgment based on characters extracted from captured image data. This page describes basic information about OCR (Optical Character Recognition) and OCV (Optical Character Verification) and the basic principle of character inspection.

Character strings are indispensable for showing information such as expiration dates for food/medicine and part numbers of assembly parts in manufacturing. They are also essential for sales order management. These strings are vitally important for management; however, management of characters can be a lot of trouble.

For example, assume that you manage orders and inventory on paper in handwriting. Management through handwriting may cause mistakes or omissions. If the book is lost, there might be no way to retrieve the orders and inventory.

In addition to preventing human errors, image processing is increasingly used to make management more reliable and reduce labor and cost. Replacing the recognition, checking and management of characters performed by human hands with an image processing system brings the following advantages:

Preventing mistakes and omissions in books
In character inspection, the captured characters are compared with those in the registered dictionary (character font groups) and each is recognized as the character with the closest shape. The recognized characters are compared with the judgment string and the pass/fail output of the final inspection is produced. It is also possible to store the output judgment result on a server or other storage device. This type of reliable procedure prevents mistakes and omissions that may occur with management through handwriting.

Reliable information management
Character inspection using image processing technology allows operators to upload the recognized character information to a server or other device, so there are no worries about mistakes such as writing the wrong characters. Product numbers or other items that consist of many characters are often written incorrectly. Handling them as digital information can prevent problems.

Thorough management of product numbers and other information also allows tracking of parts or products and collecting them quickly in case of the outflow of defective products or a product recall. It is also possible to track problems upstream to identify a problematic process to improve operations. Reliable information management is also effective for ensuring traceability.

Simultaneous reading along with barcodes or 2D codes
In recent years, barcodes or 2D codes have been introduced for product management. To read barcodes or 2D codes only, you can use a barcode reader or a handheld mobile computer. In many cases, however, an expiration date is also marked on the label along with the barcode or 2D code. Character inspection can recognize both codes and characters simultaneously. This improves efficiency dramatically because you can read all data at once and manage them together instead of reading each type of data separately.

100% inspection with minimum labor
The most effective method for preventing the outflow of defective workpieces is 100% inspection. However, checking characters visually not only takes time and labor but also involves a risk of oversight. Using image processing enables reliable inline character inspection to achieve both quality assurance and reduced labor costs. It also improves the yield rate in the inspection process and is highly effective for increasing production efficiency.

There are various types of character inspection: presence of marking, marking quality, optical character verification (OCV) and optical character recognition (OCR).

Presence of marking:

An inspection to detect the presence or absence of marking.

Marking quality:

An inspection to detect illegible characters.

OCV:

An inspection to compare marked characters and a judgment string and make pass/fail judgment.

OCR:

An inspection to read marked characters and output them for verification or control.

Although the purposes are different, the basic operation of character inspection is to extract characters one by one from a captured image, compare them with characters in the registered dictionary, and identify characters.

1. Character extraction

Extract characters one by one from the captured data.

2. Comparison against dictionary

Each of the extracted characters is compared with registered characters for identification.

In the past, there were complaints about the low accuracy of character recognition. Through various technological innovations, the accuracy of character recognition has improved significantly.

Sub-pattern registration
Registering several patterns for one character improves recognition stability. By registering various fonts as sub-patterns, you can prevent failed recognition due to different font types.

Automatic fitting
Automatic fitting is an automatic fine adjustment of characters to allow extraction in optimum conditions. Since slight variations in the status among characters are corrected individually before verification, the correlation level is increased.