Computer vision reading on stickers and direct part marking on horticultural products : challenges and possible solutions

Abstract

Traceability of products from production to the consumer has led to a technological advancement in product identification. There has been development from the use of traditional one-dimensional barcodes (EAN-13, Code 128, etc.) to 2D (two-dimensional) barcodes such as QR (Quick Response) and Data Matrix codes. Over the last two decades there has been an increased use of Radio Frequency Identification (RFID) and Direct Part Marking (DPM) using lasers for product identification in agriculture. However, in agriculture there are still considerable challenges to adopting barcodes, RFID and DPM technologies, unlike in industry where these technologies have been very successful.

This study was divided into three main objectives. Firstly, determination of the effect of speed, dirt, moisture and bar width on barcode detection was carried out both in the laboratory and a flower producing company, Brandkamp GmbH. This study developed algorithms for automation and detection of Code 128 barcodes under rough production conditions. Secondly, investigations were carried out on the effect of low laser marking energy on barcode size, print growth, colour and contrast on decoding 2D Data Matrix codes printed directly on apples. Three different apple varieties (Golden Delicious, Kanzi and Red Jonaprince) were marked with various levels of energy and different barcode sizes. Image processing using Halcon 11.0.1 (MvTec) was used to evaluate the markings on the apples. Finally, the third objective was to evaluate both algorithms for 1D and 2D barcodes.

According to the results, increasing the speed and angle of inclination of the barcode decreased barcode recognition. Also, increasing the dirt on the surface of the barcode resulted in decreasing the successful detection of those barcodes. However, there was 100% detection of the Code 128 barcode at the company’s production speed (0.15 m/s) with the proposed algorithm. Overall, the results from the company showed that the image-based system has a future prospect for automation in horticultural production systems. It overcomes the problem of using laser barcode readers.

The results for apples showed that laser energy, barcode size, print growth, type of product, contrast between the markings and the colour of the products, the inertia of the laser system and the days of storage all singularly or in combination with each other influence the readability of laser Data Matrix codes and implementation on apples. There was poor detection of the Data Matrix code on Kanzi and Red Jonaprince due to the poor contrast between the markings on their skins. The proposed algorithm is currently working successfully on Golden Delicious with 100% detection for 10 days using energy 0.108 J mm-2 and a barcode size of 10 × 10 mm2. This shows that there is a future prospect of not only marking barcodes on apples but also on other agricultural products for real time production.