WO2018154562A1

WO2018154562A1 - A system and method for the inspection and detection of coating defects

Info

Publication number: WO2018154562A1
Application number: PCT/IL2018/050166
Authority: WO
Inventors: Erez AHARONOV; Ephraim Pinsky
Original assignee: D.I.R. Technologies (Detection Ir) Ltd.; Rafael Advanced Defense Systems Ltd.
Priority date: 2017-02-23
Filing date: 2018-02-13
Publication date: 2018-08-30
Also published as: IL250756A0

Abstract

A coating-defect inspection and detection system for coated articles, comprises: a) at least one light source; b) at least one image-capturing device; and c) a graphic processing unit in which the beam of said light source and the field of view of said image-capturing device overlap in the inspection area, and are aimed toward said products from different directions; and in which said image-capturing device communicates with said graphic processing unit, which is suitable to analyze the quality of the coating and/ or surface and/ or shape appearing in the captured images by the light-and-shadow effects created by said light source.

Description

A SYSTEM AND METHOD FOR THE INSPECTION AND

DETECTION OF COATING DEFECTS

Field of the Invention

The present invention relates to a system and method for the detection of coating defects during the manufacture of pills, pellets, tablets, and the like.

Background of the Invention

The use of systems for automatic quality control inspection during the manufacturing process of pills, pellets, tablets, and the like (hereinafter collectively referred to as "pills", for the sake of brevity) is common. The purpose of such systems is to assure that the product is not defective and meets the quality control requirements.

In the pharmaceutical field, it is highly important to examine the coating quality of the product articles, since a defective item can cause a patient to miss a dose and, in some cases, such as for enteric coatings, may even adversely affect the release of the drug. Pill manufacturers are required to inspect their products and make sure that they meet the required criteria, such as a high-quality coating that is smooth, unpeeled, uniform, and flawless.

It is an object of the present invention to provide a method and system for the detection of coating defects in pills.

It is another object of the invention to provide a simple and cost- effective system for the inspection of coating in a fully automated manner and without the need for human intervention.

All the above and other characteristics and advantages of the invention will become apparent as the description proceeds. Summary of the Invention

In one aspect, the invention is directed to a coating-defect inspection and detection system for coated articles, comprising:

a) at least one light source;

b) at least one image-capturing device; and

c) a graphic processing unit;

wherein the beam of said light source and the field of view of said image-capturing device overlap in the inspection area, and are aimed toward said products from different directions;

and wherein said image-capturing device communicates with said graphic processing unit, which is suitable to analyze the quality of the coating and / or surface and / or shape appearing in the captured images by the light-and-shadow effects created by said light source.

The graphic processing unit may be implemented as a software or may be implemented as a hardware, e.g., it may be a commercial GPU or a custom-made system.

According to an embodiment of the invention wherein the light originating from the light source grazes the inspected products. According to another embodiment of the invention the light source is located essentially laterally in relation to the inspected objects. According to a further embodiment of the invention two or more light sources are located around the inspected products and they may operate alternately.

In a system according to one embodiment of the invention the image- capturing device is located at an angle that allows the capturing of an image that shows shadows that are created on the surface of inspected products as a result of the presence of pits and/ or protrusions. According to an embodiment of the invention the system is provided with more than one pair of a light source and an image- capturing devices. According to an embodiment of the invention the axes of the image-capturing device and of the light source are substantially orthogonal.

The system according to the invention may further comprise conveying elements suitable to convey the coated articles and a directing element suitable to impart a direction to said coated articles. Such conveying elements may comprise, for instance, a conveyor belt or feeding assemblies such as a slide, optionally provided with vibration-generating equipment. The directing element is configured to direct articles that go through it into a desired scanning position. According to an embodiment of the invention the directing element is shaped in order to prevent articles that go through it from aggregating with additional articles.

In a system according to one embodiment of the invention the graphic processing unit is connected to the image-capturing device. According to another embodiment of the invention the graphic processing unit and the image-capturing device are remotely connected.

The invention is further directed to a method for coating-defect detection system for coated articles, comprising:

a) providing at least one light source;

b) providing at least one image-capturing device;

c) providing a graphic processing unit;

d) positioning the beam of said light source and the field of view of said image-capturing device such that they overlap in the inspection area, and are aimed toward said products from different directions; wherein said image-capturing device communicates with said graphic processing unit, which analyzes the quality of the coating and / or surface and / or shape appearing in the captured images by the light-and-shadow effects created by said light source.

According to an embodiment of the invention the analysis of the image is performed using neural networks. Furthermore, the analysis may employ a deep learning algorithm.

Brief Description of the Drawings

In the drawings:

Fig. 1 is a schematic perspective view of the system according to one embodiment of the present invention, where the system operates on a production line, comprising one camera and one light source;

Fig. 2 is an enlarged perspective view of the system of Fig. 1, focusing on the area of overlap between the beam of the light source and the area captured by the camera's field of view;

Fig. 3 is a photo of exemplary pills with a high coating quality; Fig. 4 is a photo of exemplary pills with a low coating quality; and

Figs. 5 and 6 are photos of exemplary pills with a very low coating quality.

Detailed Description of the Invention

While the invention is not limited to any particular product, for the sake of brevity and clarity, the invention will be illustrated hereinafter with reference to the inspection of the coating quality of pills, it being understood that the invention is not limited to pharmaceutical products, and can be used for the inspection of coating quality of different objects. Fig. 1 is a schematic perspective view of the system according to one embodiment of the invention, which comprises a camera 101 and a light source 102, whose light grazes the pills moving through the system. Camera 101 can be replaced with any other image-capturing equipment. The system has the ability to inspect the coating quality of products, as will be described in detail over the following description, such as pill 103, as they move along the product line, in this example on a conveyor belt 104 or a slide, such as a vibrating slide. The system also comprises a directing element 105 that, in this illustrative example, is shaped in order to force pills that move along belt 104 to be positioned at a certain position with relation to one or more camera(s) 101 and illumination from light source 102. Another purpose of directing element 105 is to prevent pills from aggregate with adjacent pills. In case of the production of pills shaped as pill 103, for example, one desired scanning position is with their wider surface toward camera 101. The phrase "scanning position" refers to a position of the pill that allows the system to produce an image that can be used for the analysis in the coating defect inspection process.

It should be noted that there could be multiple suitable scanning positions for some products, depending on their geometry. Moreover, directing element 105 can be shaped differently for different inspected products. In the embodiment of Fig. 1, directing element 105 is an inclined surface with its lower edge 106 at a distance from conveyor belt 104, which is smaller than the largest dimension of pill 103, thus forcing it to be positioned with its wider surface toward camera 101.

According to this embodiment of the invention, light source 102 is located essentially at the same lateral plane of belt 104, at nearly 90 degrees to its direction of motion. Pills are inspected in the area of overlap between the light beam 107 and the captured area 108 of camera 101. Light source 102 is directed toward the pills on belt 104, such as inspected pill 109. At such an angle, if the surface of the inspected pills contains protrusions, they will cast a shadow that will be seen in the image captured by camera 101. In addition, if the surface of the pills contains pits, light would not reach the bottom of their surface, which can also be seen in the captured image. Examples of such images are shown in Figs. 3-6.

Fig. 2 is an enlarged perspective view of the system of Fig. 1, focusing on the area of overlap between light beam 107 of light source 102 and the field of view 108 of camera 101 (not shown in this figure). Camera 101 is located above beam 107 so it can capture images of illuminated pill for coating inspection. It should be noted that according to other embodiments of the invention (not shown in the figures) the system can comprise additional light sources and/or cameras at other locations in order to inspect products from different angles. The invention is not limited to a perpendicular positioning relationship between the beams of the light source and the camera, as long as there is an overlap between said beams. Moreover, light sources located at opposite sides can be alternated so as to obtain different images of the same pill.

As shown in Fig. 2, pills go through directing element 105 and reach the overlapping area between beam 107 and field of view 108 at a scanning position. After an image is captured, it is sent to a graphic processing unit (which may be a commercial GPU or other hardware or software) for defect analysis. In Fig. 1, camera 102 is connected to GPU 1 10, but camera 102 can also send the image to a remote GPU. In order to increase the performance of the system the camera can be a high-resolution camera that captures a high number of frames per second. The graphic processing unit used in the invention can analyze the outlines of a product and determine if they depart from the expected shape. The use of a light source is required for analyzing the quality of the coating of a product. If the GPU detects shadows on the surface, it can indicate the presence of protrusions or pits in the product. The GPU can also classify the quality of the coating according to predetermined criteria, for example, Fig. 3 is a photo of exemplary pills with a relatively smooth surface 301, thus defined as pills with a high quality of coating. Fig. 4 is a photo of exemplary pills with a relatively rough surface 401, thus defined as pills with a low quality of coating. Fig. 5 is a photo of exemplary pills with pits 501 , and Fig. 6 is a photo of exemplary pills with protrusions 601, which are defined as two cases of very low quality of coating. As an example the levels of coating are divided into three levels in those figures for the purpose of illustration, and a user can choose how to categorize different levels and to how many levels he wishes to classify the inspected products.

The images acquired by the system of the invention can be analyzed by a variety of methods. For instance, darker areas can be used to determine coating quality by a simple image analysis, using image processing methods known in the art, which are not discussed herein, for the sake of brevity. If more sophisticated determinations are desired, neural network and deep learning methods can be employed, such as the illustrative classification algorithm described hereinafter.

Classification Algorithm

The classification algorithm is based on a deep convolutional neural network (CNN) which is trained using images of good products and different examples of defected products. The training examples include recordings of machine run with good products, machine run with defected products and several types of simulations. "Deep Learning" (also called deep structured learning, hierarchical learning or deep machine learning) is a family of machine learning algorithms, which is characterised by multiple processing layers. Since 2012, these methods have gained very high popularity in the field of computer vision and in other fields as well. Currently, the state of the art in many computer vision tasks is achieved with algorithms that are based on these methods.

The advantages of Deep Learning methods are:

1. They can generate very complex functions.

2. The classifiers can learn from examples without the need for manually engineered features.

3. They can be run and trained very efficiently on parallel computing platforms such as GPUs (graphical processing units) .

The system can be trained to identify the various types of coating defects, as well as other defects such as broken pills. An illustrative description of a training procedure is provided below.

Training procedure

The training procedure according to an embodiment of the invention comprises the following steps:

1. Record good coated products.

2. Record defected Bad coated products

3. Record defected Very Bad coated products Activate Foreground- Background segmentation to detect the pixels that contain products. The segmentation generates blobs that contain the product's pixels.

4. All the blobs will be used for training unless overlapping of products or products was partially captured or was out of focus.

5. Augmentation may be applied to generate coating defects (for example "orange peel" defects) for the Very Bad coated group. The simulations generate any desired example of defects, including hard-to-detect examples (i.e., cases in which it is hard to distinguish between a correct product and a defective product because they only slightly differ in surface characteristics). Train a deep convolutional neural network to classify each frame (or each series of consequent frames) in the training set as either Good \ Bad \ Very Bad.

In this exemplary embodiments the network architecture used was that described in, Szegedy, Christian, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. "Going deeper with convolutions." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1-9. 2015, although other architectures may be used. The classification layers were replaced (this architecture contains 3 classification layers in the training phase. Only the last classification layer is used in the testing phase) with new classification layers which contain 3 neurons matching the 3 classes (Good \ Bad \ Very Bad). The network coefficients (excluding the coefficients of the classification layers) were initialized with a network that was trained on ImageNet classification challenge (as described in Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg and Li Fei-Fei. ImageNet Large Scale Visual Recognition Challenge. IJCV, 2015). The classification layers coefficients were initialized with a random Gaussian distribution. The back-propagation algorithm was used to train the network, as described in "Pattern Classification" by Duda, Hart, Stork, Wiley & Sons Inc., 2001, pp.288. All the above description of preferred embodiments have been provided for the purpose of illustration and are not intended to limit the invention in any way. Many modifications can be provided. For instance, different numbers of cameras and/or light sources can be used, concurrently or alternating, different lighting angles can be used for the grazing light, all without exceeding the scope of the invention as defined in the appended claims.

Claims

1. A coating-defect inspection and detection system for coated articles, comprising:

d) at least one light source;

e) at least one image-capturing device; and

f) a graphic processing unit;

2. A system according to claim 1, wherein the light source grazes the inspected products.

3. A system according to claim 1, wherein the light source is located essentially laterally in relation to the inspected objects.

4. A system according to claim 1, wherein the image-capturing device is located at an angle that allows the capturing of an image that shows shadows that are created on the surface of inspected products as a result of the presence of pits and/ or protrusions.

5. A system according to claim 1, wherein two light sources are located around the inspected products.

6. A system according to claim 5, wherein the light sources operate alternately.

7. A system according to claim 1, wherein there is more than one pair of a light source and an image-capturing devices.

8. A system according to claim 1, further comprising conveying elements suitable to convey the coated articles and a directing element suitable to impart a direction to said coated articles.

9. A system according to claim 8, wherein the directing element directs articles that go through it into a desired scanning position.

10. A system according to claim 8, wherein the directing element is shaped in order to prevent articles that go through it from aggregating with additional articles.

1 1. A system according to claim 1 , wherein the graphic processing unit is connected to the image-capturing device.

12. A system according to claim 1 1, wherein the graphic processing unit and the image-capturing device are remotely connected.

13. A system according to claim 1, wherein the axes of the image-capturing device and of the light source are substantially orthogonal.

14. A method for coating-defect detection system for coated articles, comprising:

e) providing at least one light source;

f) providing at least one image-capturing device;

g) providing a graphic processing unit;

h) positioning the beam of said light source and the field of view of said image-capturing device such that they overlap in the inspection area, and are aimed toward said products from different directions;

wherein said image-capturing device communicates with said graphic processing unit, which analyzes the quality of the coating and / or surface and / or shape appearing in the captured images by the light-and-shadow effects created by said light source.

15. A method according to claim 14, wherein the analysis of the image is performed using neural networks.

16. A method according to claim 15, wherein the analysis employs a deep learning algorithm.