China is a big country for fruit production and consumption. China's fruits are not only rich in varieties, but also foods that use fruits as raw materials, such as canned food and jellies, are also quite large. However, foreign materials such as hair, fiber yarn, paper scrap, metal, and paint may be inadvertently mixed in the processing of the fruit fruit, thereby adversely affecting product quality and consumer psychology. At present, most food production companies still use manual naked eye to detect whether the product in process is contaminated with foreign matter, and have the disadvantages of low efficiency, high absent detection rate, and large labor volume.

According to the report of the "Journal of Agricultural Engineering", Chinese scientists have carried out research on image monitoring of various fruit and fruit materials that have been cut into chunks before filling on a jelly or canning production line, and automatic determination of the presence or absence of foreign materials has been applied, and achieved certain results.

When people eat or even see foreign objects in food are always disgusting, there are occasional incidents of claims to sales and manufacturers. In order to ensure that there are no foreign materials in food, manufacturers need to set up multi-channel inspection stations in production. The vast majority are manual naked eye inspections. Human eyes and hands cooperate with a high degree of intelligence and flexibility, and can recognize and propose a variety of foreign body defects. However, visual fatigue, physiology and subjective factors will bring about differences in work quality and inefficiency. The use of machine vision technology to replace manual inspections is the development trend of modern production.

With the requirements of improving product quality and increasing labor costs, companies are eager to apply machine vision technology to achieve automated detection of industrial production. However, in terms of quality of agricultural products and quality of food processing, the original research results at home and abroad mainly focus on the detection and classification of size, shape, maturity, surface damage and defects, etc., of intact and relatively dry fruit, and in the detection of foreign matter. Only research on the detection of certain foreign materials on a single species of fruit material, such as orange pie, has been conducted.

In the processing of products such as canned goods and jellies, in order to facilitate filling, the flesh is generally divided into blocks. However, the shapes and sizes of various fruits are different, and the shapes and sizes of foreign bodies are also varied. For example, hair and filaments are slender. Paints, metal shavings, etc. are lumps; various varieties of fruit are of various colors, such as light yellow for apple, dark yellow for orange, and white for coconut, while the color of various foreign materials is also varied, such as black hair, paint, and filaments. Mostly colored, scrap iron is silver or black. The difference in scale and chroma between various foreign objects and flesh is very different. These features have brought great challenges to the automatic identification of foreign objects.

The author of the article “Method for identification of multi-type foreign bodies based on machine vision” developed a set of multi-species, multi-species, and various foreign objects that may appear on wet reflective flesh. Type of foreign body automatic detection system. The mechanical device is used to automatically arrange the fruit layers on the conveyor belt in an automatic manner. The industrial camera installed in a proper position monitors and photographs the conveyed fruit, and the collected fruit images are transferred to a computer and processed by the image processing software. Analyze and judge. According to the characteristics of differences in the color and brightness of fruit and foreign objects, the fruit varieties of each variety are divided into two categories, and different image processing strategies are used to identify foreign objects.

Fruits rich in color such as yellow peaches, pineapples, etc. are separated and recognized according to the color of the flesh and foreign objects; foreign objects are identified based on the edge contours of foreign objects for flesh with a white or transparent color such as coconut and gelatin. After a large number of tests and verifications, the system can effectively detect many types of foreign materials on the multi-variety fruit conveying line and remove fruit containing foreign materials, providing technical support for automated production and testing of enterprises.

This research report was published in the 3rd issue of the Journal of Agricultural Engineering, 2011, entitled “Methods for Identifying Foreign Types of Pulp Based on Machine Vision”. The first author was Prof. Yanming Yan from School of Mechanical and Automotive Engineering, South China University of Technology.