Machine Vision Applications in Plastics Injection Molding

In the manufacture of consumer products, quality control is of utmost importance. In most instances where manual processes are involved, the individual who operates the machinery also performs this task. In injection molding, for instance, this means ensuring the correct orientation and quality of print on the manufactured parts and components. In manually operated equipment, this inspection process presents a challenging task because often, after prolonged hours of work, fatigued operators will insert labels inside the molding dies in improper orientation, resulting in rejected parts. In other cases, because of the heat in the molding process, if there is excess ink on the labeling equipment, the prints may be smeared or distorted. Machine operators occasionally miss or overlook quality issues. To alleviate this inspection and operation problem, it is now possible to incorporate vision systems which will ease the inspection task for the operators, allowing them to concentrate more on the manufacturing process. Vision systems provide means by which continuous and autonomous inspection can be achieved during production.

Computer vision techniques are increasingly being used as alternative methods to conventional inspection and monitoring applications, due to their simplicity and ease of set-up, relative insensitivity to ambient noise, noninvasive means of gathering information on objects without contact, marking or specimen preparations, and the potential for online applicability.1"3 Machine vision has continued to play a major role in the integration of automated manufacturing, both from a quality and inspection perspective to more advanced applications such as motion control and robot guidance. For example casting mould,mold making,plastic injection mold etc.The role of machine vision in providing solutions to manufacturing automation has been recognized by both researchers and engineers in industry.

In the plastics manufacturing industry, vision applications have been used mainly to inspect the quality of molded parts, especially for missing features or badly formed sections. Although fairly automated, the plastics molding industry still has a large component of its operations carried out manually, especially for medium- and low-volume rates of production. For instance, the placement of print template or metallic parts to be included in the finished product is still done manually. Under such circumstances, it is possible to have a wrongly inserted print template which would result in the rejection of the part.

In some cases, due to process variations, a particular plastic part may posses a smudged or faint print. These latter quality problems are often not easy to detect in the processing of plastics. In this paper, the application of a smart vision system in the automatic detection of poor or improperly oriented print templates is presented. The method used to detect improper orientation is based on edge detection and feature identification. More advanced algorithms like segmentation, template matching, and character recognition are utilized for ensuring proper print quality.

To achieve machine vision successfully, and implement the use of computer software, the viewing area must be represented in digital form.The purpose of image acquisition therefore is to capture the optical data and change it to a form that will facilitate convenient and efficient processing using a computer. An image is typically embedded within a viewing area covered by the video capturing mechanism or sensor. The most common type of capturing mechanism is the charged coupled device (CCD). When a light source hits an object, it is reflected to the sensor through appropriate lenses. The photons cause electrical charges to be created in the CCl), thus generating analog signals on a 2-D array. The intensity of the charge at each discrete point in the 2-D array is proportional to the photon energy impinging on that point, determining the brightness or intensity of the light. Therefore a typical optical image system is a continuous 2-D function, h(x,y) whose value at any pair of spatial coordinates represented by the Cartesian coordinate system (x,y) is the intensity of light at that point. The continuous function h(x,y) must be quantized (or digitized) so that it can be easily processed by computer.

The most common method of digitization is a combination of spatial and amplitude quantization in which the viewing area is divided into a matrix of m by n cells or pixels: m and ç are integers. The image is sampled at these discrete points in the viewing area. Each pixel is then assigned a numerical value that is a digital representation of the initial analog value. h(x,y). through the analog-to-digital (A/D) conversion system.The numerical value will depend on the bit resolution of the A/D converter which in turn determines the number of gray-scale levels in which the image can be represented. Therefore the new image data, I(m.n). will have values between O (dark or lowest intensity) and 2^sup r^-l. where r is the bit resolution of the A/D converter.The acquired image must then be filtered to remove any noise, and enhanced for analysis.This stage is known as processing. There are five common types of operational approaches for primary processing of pixels.