The applications presented utilized tools that are developed from image processing algorithms. In the inspection of correctly inserted print templates.it has been shown for instance that two feature detection algorithms can be applied; in this case canned in the software described as feature detection tools. If the print template is inverted in any way, whether upside down or left to right, either or both tools will return a "fail" based on the prescribed dark or bright feature size required across the tools. For print quality, two methods have been described. The first is to use a template match of a good print and compare it with other parts. This was mainly used to detect smudges, smears and poorly printed characters. Provided the degree of mismatch is adequately defined, the template match can be used fairly adequately. In this particular case study, the template match could not be tested extensively because of lack of adequate samples.The second method used algorithms for reading optical characters (OCR).The printed characters on a good plastic part, which are not standard OCR fonts, were read into an OCR tool.Through software, the tool was trained to recognize these as OCR fonts with varying degrees of acceptance. Like in the case of the template match, the OCR was not tested extensively due to lack of adequate different production samples.For example casting mould,mold making,plastic injection mold etc.

A variety of such software and hardware exist in the market today. The comparison between the different software/hardware platforms is not intended to be the subject of this paper; however a comprehensive listing can be obtained from the Automated Imaging Association.20 Plastic molding processes are widely used in the manufacturing of various engineering and consumer items. The growth of the plastics sector has seen a slight decline (-5% overall) in the U.S. since 2000 due to the increasing costs of fuel and gas, the weakening of the dollar against major currencies in the world, and more so, the movement of manufacturing to Asia (especially China). This deficit has been absorbed mainly by China, Canada and Japan. Despite this, there is still a substantial proportion of manufacturing companies in the U.S., especially in the molding industry. Thus there is still a great need for improved process and quality control. This paper presents a simple approach that utilizes commercially available hardware and software for machine vision applications to automate the inspection of molded plastics. Generally, the training required for using such systems is minimal since most software packages supplied by vision systems manufacturers are user-friendly. The inspection for quality also requires very simple tools like those that have been demonstrated such as feature count and template matching.

After the molding process is over, the part is removed from the mold cavity manually, and visually inspected for quality. Despite this, process variations could cause minor blemishes or smears on the print that are not immediately visible to the operator. Figure S shows an example of such a defect with a close-up on a print revealing a small smear on the letter "d." Two methods can be used for this inspection.The first is to use a temporal operation such as the template match described in section 2.S.The training data is obtained from a captured image of what is perceived as a very good quality print. Subsequent images can then be captured from parts as they flow along a conveyor, and each image compared with the trained data. A problem such as a smear or a missing character may cause a mismatch in the number and position of dark pixels that are in the image. Figure 6 is an illustration of the application of this tool.

Another useful tool that would he used to inspect a print is an optical character reader (OCR). Although the prints are not true optical characters, using the software, normal non-OCR font characters can be trained to correspond to a particular print image.After an image of a good print is captured, using this software, the actual character string is typed into the OCR reader. The reader is then "trained" to interpret the image data as corresponding to character string from the keyboard entry. After several trials, an acceptance level for allowing the captured image characters as ones that match the corresponding keyboard characters is determined. If any of the characters from a subsequent part contains a large smear it would not match the trained data set. Additionally, if there is a missing character on a plastic part, the string will not match the trained one. An example of this is shown in Figure . There are two limitations with this tool however. The first is that there ought to he adequate spacing hetween the characters for it to work effectively. secondly, minor smears on the prints may not easily he detected. Such limitations have heen addressed hy the use of advanced processing algorithms such as those that utilize neural-fuzzy classifiers.