Vision - a new technology to control leathers

Introduction As a natural product, leather still retains traces of the original animal, giving it a unique character. But the excellence of leather products demanded by the customer requires skins that meet precise criteria of quality, particularly where defects are concerned. Faced with inevitable defects, both tanners and manufacturers need to know how to optimise the potential of their raw materials. Until now everything has depended solely on the skill of the grading and cutting specialists to detect defects during the tanning process, achieve the best grading and maximise the placing and cutting of pieces for shoes or leathergoods. Together with their partner Turner, CTC have developed visual technology for the leather industry consisting of a unique system that detects and identifies defects at every stage from wet-blue through to finished leather, enabling skins to be graded and classified. This innovative, automated, accurate and objective system gives tanners and manufacturers of shoes and leathergoods valuable support in making the best use of skins. Presentation of the skins As the raw material of the tanning industry, skins have an uneven structure and non-uniform surface appearance. Skins are often divided into four quality zones: butt, shoulder, tail-end and bellies. However, skins of different origins have varying characteristics such as size, softness, size of grain etc. In addition to the influences of environmental factors such as climate, habitat and diet and the special characteristics of each individual animal such as breed, age, gender, coat and state of health, must be added damage to the skin. This affects the quality of the finished leather and, therefore, has a major influence on its market value and end use. Defects appear at three different stages: * on the live animal due to mechanical, health, parasite or genetic factors * when the skin is removed * during storage and processing The diversity of the types of defect means that users have to constantly make allowances: some defects influence the appearance of the finished leather whilst others damage its structural properties. Leather can be used for a wide range of purposes. The eventual choice depends on: * the sector for which the leather is intended (furniture, shoes, leathergoods) * the quality of the finished product (mass produced or luxury quality) * the quality of the leather itself (choice) The position of the defect must be taken into account as well as the possibility of integrating it into the lasted margins during manufacture of the finished product. Thus, advance automatic detection of defects is of strategic benefit for both tanners, allowing them to check quality and optimise the process for each individual skin in terms of its defects, and manufacturers of finished leathergoods, enabling them to perform quantitative and qualitative grading and maximise potential surface area when cutting. Objectives of the project In order to meet the high expectations of the sector, CTC have developed a system that meets several objectives: * Capturing a high quality image of skins * Detecting defects at each stage of leather manufacture from wet-blue through dyed crust to finished leather * Classifying defects by type, seriousness and size Description of the system The system comprises three components: the conveyor unit, the image capture system and the image processing software. Conveyor unit The prototype is based around a belt inclined at an angle of 45°. This allows skins to be fed onto the system and displayed in a small space. The belt is silicone coated to enable it to resist the harsh environment of a tannery and can support skins that are both large and heavy (eg calfskins). The skins are conveyed towards the capture system located at the top of the unit. This configuration of the system can be modified to suit particular sectors: tanning or shoe and leathergoods manufacturers. The capture unit is, therefore, able to analyse goat, sheep and calf skins. For shoe or leathergoods work the capture station can be positioned close to the production line or incorporated into the cutting installations. Image capture Feeding is by means of the translucent belt. The skin is held flat by a hinged strip that presses the leather flat before an image is captured by the video camera. The detection system includes a 6,144 pixel video camera with a 35 mm lens. The captured image is converted to 1,024 grey levels. The resulting image is of excellent quality with a resolution of 0.24 mm per pixel. Good lighting is essential for the detection of defects, both for this machine and the work of leather graders and cutters. Two types of lighting are used: the first works by reflection and gives a diffuse light, showing up both structural and visual defects. The second light-source operates by transmission and is placed underneath the translucent belt. This light-source saturates those areas of the belt not covered by the skin while eliminating shadows by increasing the contrast between the skin and the belt. At this stage the exact surface area of the skin is calculated. The image capture time for a wet-blue or finished leather application of goatskin is 10 seconds and calfskin is 30 seconds. For finished leather, automatic adaptation depending on the colour and brilliance of the leather is performed and capture can be made of all types of surface (smooth, artificially or naturally grained) to give better defect visibility than with the naked eye. An algorithm using Bézier curves is used to correct the image. Image processing The core of the system is the image processing software. The algorithms applied mean that the system is capable of considerable upgrading. Image processing time for a goat skin is 10 seconds. Skins are analysed in four phases: 1 geometrical analysis: quality zones and direction of stretch (elasticity); division into areas with the same shade for wet-blue work and finished leather (marks on animal's coat or non-uniformity of shade) using a Split & Merge algorithm based on a QuadTree structure. 2 detection, location and identification of defects: this phase consists of several modules used to detect the various types of defect one after the other according to size (small, medium and large) and shape (linear or circular). The detection phase consists in sweeping the image using shapes which are similar to the defects. The algorithms are very efficient in detecting defects that are difficult to see on images that are badly blurred by the grain of the leather. They give a probability image with maximum similarity. The location phase generates the rectangular contour of the shape that gave the best probability for each defect above a user-defined threshold. Detection accuracy can, therefore, be adjusted depending on the quality required by the customer. The characterisation phase very accurately establishes the contour of the defect by using active contour algorithms in the area. The identification phase uses the geometrical and photometric parameters of each defect detected to assign it to a class. To date nine classes have been defined: wrinkles, veins, holes, parasite defects, tears, scars, crimping, colour marks and white spots. If users have particular requirements for manufacturing a product, they may eliminate defects that can be overlooked by selecting one or more classes of defect and adjusting a visibility parameter. 3 automise the cutting process: generating an image including defects and the contour of the skin uses five colours to identify the level of seriousness of a defect. This system is interfaced with a nesting software, an automatic placing software. Depending on the seriousness it may be considered acceptable in certain areas of the skins. 4 automatic grading of skins: This phase assigns a level of quality to each skin according to user-defined criteria. These criteria are used to determine the acceptable surface percentage for each category of defect. Based on these criteria and the defects detected, the system assigns a level of quality from 1 to 4 to each zone as well as to the overall skin. For example, the system classifies the butt as category 1 if the surface of the butt has less than 5% wrinkles, less than 3% parasite defects etc. This means that the butt will be classified 2 if it has between 5% and 10% wrinkles. Each part (butt, tail-end, shoulder and bellies) is given the same treatment. When a choice has been made for each part, a truth table is used to classify the skin. All these parameters can be modified by the user and set differently for different users. The grading of skins at various stages (wet-blue, crust or finished) can be fully automated. When the system has classified a skin it can be combined with an automatic stacking machine for creating batches. Conclusion This tool will enable objective sorting and improve the detection of defects at the earliest stage. Carrying out added value processes on bad quality raw materials (wet-blue leather for tanners or finished leather pieces for product manufacturers) will bring only losses. This new technology will allow the setting up of new working methods which will allow for improvement in control and profits on raw materials. It also offers a new method to facilitate communication between tanners and their clients.