TW201736831A - System and method for inspecting containers using multile images of the containers - Google Patents

Abstract

A system and method for inspecting containers by detecting reflected light from defects are disclosed. A light source illuminates the contents of a container. One or more cameras are oriented to detect and the light reflected by one or more defects contained within the container, which are reflected from two or more mirrors towards the camera or cameras.

Description

System and method for detecting containers using multiple images of each container

The present invention is generally directed to the detection of bottles and containers, and more particularly to the use of light to detect bottles or containers to indicate artifacts or commercial variations.

Bottled beverages are produced, sold and consumed daily. Since these beverages are consumer products, they are subject to strict quality control and testing requirements, and often perform inspections on containers directly on the production line. The production process includes functions such as washing the bottles, detecting the contents of the bottles, filling the bottles with beverages such as soda or beer, packaging, and labeling the bottles. The quality inspection of the filled containers occurs in the case of a single-file travel, in the case of a filler discharge, or in-feed or out of the labelling machine. When the material is out-feed.

The cost-saving method for detecting small items (for example, 1 to 3 mm) has not been realized. Fragments and other physical contaminates may lie in the base of a container and escape detection. However, the inability to detect contaminants has resulted in expensive products being recalled and damaging goodwill. Today, despite the extensive development of many testing companies (extensive development) Programs, but there are only a few effective detection solutions that can reliably detect physical contaminants such as small and any shape - many of the existing methods encounter false "positive" detections (false "positive" detections). But this adds cost.

U.S. Patent No. 7,982,868 discloses a method and apparatus for detecting foreign substances in a container using two or more different orientations, wherein the camera and the light source are interconnected such that they are short Two or more images of a container filled with liquid can be recorded in a manner that is mutually differential illumination. In this patent, it is possible to determine the probability that the substance to be tested actually represents the undesirable foreign substances. In other words, the decision to discard the container is based on probability distributions. The methods and systems in this patent may result in dragging of the bottles, and most images taken by different cameras at different locations require high data rates and camera imaging rates.

The present invention is also directed to such and other needs in order to inspect bottles and containers that always require improved methods and systems.

The present invention describes a detection system and method for detecting a bottle or container of a stolen object using a light source, such as directed light, collimated light, or scattered light. Light is reflected by the use of a plurality of lenses, or lenses are used to direct light onto the axis of the bottle or lens to capture light that is reflected or refracted by physical contaminants. Described here The system and method can provide a bottle/container detection assembly that performs multiple detection functions than the current system's smalller footprint.

One aspect of the invention pertains to a system for detecting a bottle or a container. The detecting member includes a light source that generates a light source or a light beam and a camera or a plurality of cameras that can be guided inside the bottle by a contaminant or cockroach (a vent tube is not a fragment) The portion of the light that is reflected or refracted.

10‧‧‧ camera

11‧‧‧ camera

16‧‧‧ lens

20‧‧‧Mirror

25‧‧‧Mirror

27‧‧‧beam splitter

29‧‧‧Mirror

30‧‧‧ Images

35‧‧‧ Images

37‧‧‧ Images

39‧‧‧ Images

40‧‧‧ aperture plate

45‧‧‧Light source

80‧‧‧Band

85‧‧‧Support board

88‧‧‧ conveyor belt

100‧‧‧Detection system or component

150‧‧‧ lens

200‧‧‧ bottle

D‧‧‧Distance

F‧‧‧Shopping

Figure 1 shows a detection system for detecting contaminants during detection of a container in accordance with an embodiment of the present invention.

Figure 2 shows a bottle above the light source on an aperture plate in accordance with an embodiment of the present invention.

Figure 3 shows a side view of a detection system in accordance with an embodiment of the present invention with a container moved to the inspection system and moved to a conveyor belt.

Figure 4 shows a detection system for detecting contaminants during detection of a container in accordance with an embodiment of the present invention.

Figure 5 shows a detection system for detecting contaminants during detection of a container in accordance with an embodiment of the present invention.

FIG. 6 shows a pollution detection effectiveness response curve according to an embodiment of the present invention.

Figure 7 shows two of the same bottles taken at different angles in an embodiment of the invention. image.

FIG. 8 shows a mapped image generated by a calibration map transformation according to an embodiment of the present invention.

Figure 9 shows a transaction line in accordance with one embodiment of the present invention.

Figure 10 shows a region of interest for an embodiment of the present invention.

Figure 11 shows the edge of a treatment line in accordance with an embodiment of the present invention.

Figure 12 shows the cylindrical shape of a bottle of the embodiment of the present invention.

Particular embodiments are systems and methods for detecting commercial variations or stolen goods in bottles or containers. For example, the booty may be a glass fragment produced at various product line locations of the filer and the crowner of the bottle. These shrimp mites include glass shards that may be introduced during the filling process and may fall at the lowest point of the base of the bottle. These fragments may not reflect or refract enough light to detect. False alarm features may be via mold marking, ascii text, and other embossing, glass texture, discontinuous "smile" processing (discontinuous'smiley) Face'artifact), water droplets, exterior foam bubbles, and internal glass inclusions such as foams or cracks that can be generalized by certain embodiments. The false alarm function can occur outside the bottle or inside the container wall. The stolen object or object of interest is located inside the bottle.

For details, please refer to the drawings. FIG. 1 depicts a system or inspection component 100 for detecting unacceptable commercial variations or defects during inspection of a container 200. It consists of a transparent or partially transparent material such as glass. In general, the detection system or detection component 100 is reflective or refracting by light in each container, sufficient to detect a processed article or a sputum (eg, a sediment), and is capable of discriminating the location of the deposit in a three-dimensional image. Reduce the false alarms on the outside of the bottle or in the wall of the bottle. The term "reflection" as used herein refers to the reflection and refraction of light under detection. The camera or sensor 10 is operatively coupled to a processor that determines if there is any debris in the bottle 200.

As shown in FIG. 1, the detection system or assembly 100 for detecting debris in the bottle 200 has a light source 45, a camera 10, and one or more mirrors 20, 25 from a platform. In one embodiment, the two mirrors 20, 25 are placed to maintain the same path length from the object to the camera. When the bottle 200 does not contain sputum, the light emitted by the light source 45 passes through the bottle 200 in a manner that is substantially unreflected. However, when the bottle 200 contains a smear (e.g., a shard) in the path of the ray 30, 35, the smear F reflects at least a portion of the light reflected by the camera 20 and the mirror 25 by the camera 10. Grab and/or measure. In some examples, the angle of light (from each mirror 20) is 10 to 20 degrees (from the bottle 200). In still other examples, there may be a light source 45 below the container 200 with a gap between the base of the container 200 and the light source 45. The camera can include an image sensor or a machine A processor of machine vision or visual software. Most containers or bottles 200 have mold marks on the wall of the bottle, such as mold dots, mold lines, and alpha numeric marks, which can interfere with the bottle. Imaging of internal contaminants. By combining these different images of the bottle 200, the third spatial position of the possible stolen goods can be identified by the system 100. A complex image of the same object or scene can be used to calculate the appearance of the three-dimensional shape of the object.

2 shows an embodiment of the system 100 indicating the bottle 200 located above the light source 45 (e.g., collimated light) that passes through an aperture plate 40. The collimated light passes through the aperture plate 40, and some of the contents of the bottle 200 are reflected onto the mirrors 20, 25 via the paths 30, 35 onto the lens 15 of the camera 10. The bottle 200 can be secured (e.g., secured by straps 85) at a distance D above the aperture plate 40 (also shown in Figure 3).

The distance D from the base of the bottle to the bottom edge of the aperture plate 40 depends on the physical size of the bottle, and for a bottle of 65 mm diameter, for example, is about 0.5 mm to 2 mm. The size of the pores is important, about 90% of the diameter of the bottle 200. This will cause the wall of the container to act as a light pipe, illuminating any physical contamination in the path of the light, and enabling physical discontinuities on the wall of the bottle, such as random molds. Random mold marks such as dots, alpha numeric marks, and lines, which may confuse machine vision algorithms used to detect contaminants in a bottle. The space between the base of the bottle 200 and the aperture plate 40 The gap/distance D reduces the mirroring artifacts produced by the base of the bottle and the conveying surface of the contact. Furthermore, water and detergents are often used as lubrication for the bottling line conveyor belt, but they can cause false imaging artifacts due to foam or water droplets. This void can contribute to the use of an air knife to dry the base of the bottle to improve the tolerance of contaminant detection. In addition, the void prevents dragging of the base of the bottle as it can cause the bottle to tip down during the test.

3 shows a side view of the system 100 of the present invention showing the container 200 moving past the system 100 and toward the conveyor belt 88. In this example, the moving strap 80 is positioned over the base of the container 200 to facilitate viewing of the base-related extent of the bottle, as shown by the camera image/ray paths 30 and 35 as shown in FIG. The system includes two compliant belts 80 that are constrained by a backing plate 85. The moving straps 80 can pull the bottle 200 through the system before placing the bottles back into a conveyor track 88. In an alternative embodiment (not shown), the bottle is grasped by a mechanical device such as a clamp and a starwheel or linear robotic type system is used. Mechanized delivery system.

In one example, a glass container inspection computer has a portion of a processor, the glass container inspection computer including a memory coupled to the processor, and one or more interfaces coupled to the processor and One or more input devices (such as image sensors, position sensors, user interfaces, etc.) and/or one or more output devices (such as light sources, material handlers, displays, etc.) are coupled. The The computer may further include any auxiliary devices such as a timer, an internal power supply, and the like (not shown). The processor can process the data and execute instructions that provide some of the functionality of the present invention. The term "instructions" as used herein may include, for example, control logic, computer software and/or firmware, stylized instructions, or other appropriate instructions. The memory can include any computer readable medium, or a variety of media that are installed to provide at least some data, data structures, an operating system, applications, program modules or data, and/or Or temporary storage of other computer software or computer readable instructions (which may provide at least some of the functionality of the present invention and which may be processed by the processor). The data, instructions, and the like may be stored as look-up tables, formulas, algorithms, maps, models, and/or any other suitable format.

4 and 5 show additional embodiments of the detection system 100 of the present invention. 4 shows a detection system 100 in which light 45 is reflected to a camera 10 and the lens 150 to which the camera 10 is coupled is mounted adjacent to the bottle 200 to produce a single image containing two of the relevant areas of the base of the bottle. Images 30, 35. The two images 30, 35 are obtained via two mirrors 20, 25. The two mirrors 20, 25 are arranged to produce an angle compensated image of contaminants or artifacts within the bottle. Contaminants can be separated from the processing of the bottle, while the processed product is present outside the bottle, such as mold scores and water droplets. Stereoscopic views of the stolen goods can be processed to determine if it is a commercially significant likelihood.

Figure 4 shows another camera 11 and lens 16 for use with a beam splitter 27 to capture additional portions via the base of the bottle 200. Image/ray path. The detection assembly includes a directed light source that emits light, and at least two cameras 10, 11 positioned to detect one of the light reflected by the object of the bottle, such as a fragment. Share. This arrangement allows for additional detection of certain types of artifacts that are capable of producing a stronger reflection or shadow of light as seen by the base of the bottle (if compared to images acquired by images 30, 35). Combining the images 30, 35, 37 can further increase the detection capability of the system. In another embodiment, an additional strobe light is used to side light bottle 200 and image 37 to compare dark field images that may be contaminated within the bottle.

Figure 5 shows another embodiment of the system 100 of the present invention having a third mirror 29 that is positioned to be offset from the axis of the mirrors 20, 25 to create additional images. The information in image 39 can be used to compare with the stereoscopic view produced by image/ray paths 30, 35 to provide more accurate artifact detection. This information can further provide the size and shape of the detected contaminants.

Figure 6 shows an example of a pollution detection effectiveness response curve, which summarizes the experimental results to determine the optimal angle for taking the pollutants and the like when presenting mold marks and lines. . At shallow angles, the mold marks (points and alphanumeric marks) are less effective because they are less noticeable, but the mold lines are very clear, making the overall contaminant detection rate tend to moderate. At high image angles (which are ideal for imaging contaminants or sputum), the mold marking is very noticeable, The mold line is very inconspicuous. The optimum viewing angle is therefore chosen to be approximately 20 to 25 degrees, minimizing the effect of any mold marking, with a mirror angle differential of approximately 10 degrees.

Figure 7 shows two images of the same scene or bottle 200 taken at different angles. In this example, two images at angles of about 10 and 20 degrees are displayed in the camera frame (image) and obtained using the same light source. And at the same time (no relative motion). Paired images are combined and analyzed to determine if the bottle contains sputum. The idea is to use a set of mirrors, a set of cameras, or a combination of them, using stereo vision, based on 3D position and 3D morphological attributes, from other bottle features and processed products, To distinguish between foreign bodies. The spatial position and orientation of the camera is determined using prior art imaging calibration and fitting procedures. For the image of each pair of targets under test, the object fitting algorithm is manipulated to identify and correlate the target in two contexts and derive the 3D position of the target (eg, use The calibration mode of the prior art). For example, the wall portion of a glass can be mock-applied using the curvature of the semi-circular processed product "smile-finished product" at the heel and wall of the bottle. This processed product describes the curved surface of the unique glass wall of each image. It is important to note that the dimensions of the individual bottles are different - especially when passing through different molds and the mold is filled with different amounts of glass material, the thickness of the base of the bottle and the diameter of the bottle (elliptical).

Preferably, the spacing between the camera, the object, and the image is minimized to a large aperture to increase the angular coverage of the reflected illumination from the debris. can Two or more cameras that are independent of each other are placed at different angles to form an image. However, the physical size of the camera and lens can limit the angle of separation or standoff-distance between images. The one is slightly bigger. A smaller angle (1 to 5 degrees) indicates insufficient depth resolution for the application being tested. Large spacings (such as 25 to 30 degrees in a vertical plane) can cause reflection losses on one of the images of the target, making the object fitting complex. This preferred angle is different for some embodiments in the horizontal plane and more symmetry in optical illumination. This is done by two cameras spaced about 5 and about 25 degrees apart, resulting in 3D reconstructions, and initially investigated.

The optical distortion characteristics of the lens are compensated by calculation, and the spatial position and direction of the camera are also the same. An imaging calibration phantom is placed in the detection zone, the phantom having various high-contrast features at known locations, which are used to correctly calculate the position of the imaging camera relative to the detection target With direction. This is used as reference data to process subsequent images and generate 3D data. When the camera or mirror is actually moved, the calculation can be done again.

The detection member 100 can include two or more mirrors, and Figure 1 shows two mirrors, and two or more mirrors or cameras can be used to create an image of the bottle 200.

In yet another embodiment, an extension of the present technology includes utilizing data from four cameras in addition to using two scenes per camera. The four cameras produce "smiley" finished products from different and overlapping, which will allow for a more accurate bottle wall position when comparing only one camera. Additional mirrors in the horizontal plane can generate additional solid mirror image data that can be used to improve 3D accuracy.

In yet another embodiment, two cameras are used instead of two images produced by two mirrors and one camera.

In still another embodiment, the optimizer of the sensing member 100 of the present invention can include determining the artifact of the bottle 200 sufficient to reflect at least a portion of the light onto the camera 10. Therefore, more than one camera 10 can be used. As shown, the cameras 10 can be positioned in a single plane of a bottle 200 to produce a solid mirror image. The cameras 10 can be positioned at different angles relative to the respective camera 10 and the bottle without departing from the scope of the present invention.

Furthermore, camera positioning, illumination, and stereo match can be adjusted and optimized. Camera positioning (x, y, z) and camera lens calibration can be calibrated to capture reflected light. The level and direction of light illumination can be adjusted to produce an appropriate signal. For example, a back-lit or side-lit view is an extension of this embodiment. A sufficiently solid and correct stereo match can result in improved detection and also allows for discriminatory of false alarms.

In yet another embodiment, the sensing member 100 can include reflective structures that can reflect and focus portions of the light reflections, while the reflective portions of the light can combine the two reflective structures before reaching the camera. The detecting member may also include a transportable The device that delivers the bottle across the light source is approximately 300 mm in length. Additionally, the detection member can have one or more illumination sources, which can be continuous, scintillating, or a combination thereof. Furthermore, the detecting member can include support strips that can guide the bottle into a detection position. Additionally, the detection member can include a conveyor system having a length of less than about 1,200 mm.

Additional cameras can be positioned within the detection component to perform other detections such as fill level detection, floating target and sink target detection, and bubble detection. In some examples, the camera of the detection member can be offset from the horizontal by about 20 degrees or about 10 degrees, and/or offset from the optical axis by about 70 or 80 degrees. In yet another example, the sensing member can be offset from the horizontal by from about 10 to about 20 degrees and from about 70 to about 80 degrees from a ray axis. The diameter of the light is substantially equal to the inner diameter of the bottle.

In another embodiment, the system can use an image obtained by utilizing a position trigger on the bottle transport line to cause the light source to blink for a short period of time and capture a single image with the camera. In some examples, camera 10 has a frame rate of no more than 50 frames per second, which is a typical high speed bottle conveyor that reduces the cost of the overall system.

Although the present invention can utilize a camera in the most simplified implementation, the use of additional cameras and mirrors can cover a relatively large range of bottle bases and, in some bottle formats, can also increase contaminant detection. Reliability. An alternative to the use of two mirrors is the use of two direct scene cameras, which are costly but meet the high data processing requirements.

A further embodiment of the invention (see Figures 1 to 5) uses two or more collimated light sources 45 on a plate 40 having corresponding additional cameras 10 mounted in accordance with each light source 45. , the lens 15, and the mirrors 20, 25, and in conjunction with moving the two conveyor belts 80 at different speeds, when the bottle moves over the plate 40, the bottle is rotated, and the positioning of the bottles at each of the light sources 45 is different. , through the various imaging systems, there are more detection coverage areas in the bottle.

Still another embodiment of the invention includes the use of light over a bottle to conserve contaminants (partially floating or completely floating), determine the fill level, and measure the degree of foaming.

The term "bottle" or "container" is used to include a transparent or translucent container. In some cases, the bottles are tinted bottles, and or dark liquids, and the wavelength of the illumination must be near the infrared portion of the spectrum. In addition, polarized light can be used to detect patterns of certain visible contaminants, such as cellophane or to handle very reflective containers.

The detecting member 100 may further include reflective structures that reflect and focus the light reflecting portion when the light reflecting portion moves from the bottle to a camera. The sensing member also includes reflective structures that are positioned to reflect and focus the reflected portion of the light before the reflected portion of the light reaches a camera lens. As shown, the reflective structures have a planar structure of planar surfaces. However, there are reflective structures that can utilize other geometric configurations rather than planar surfaces such as convex and concave surfaces. The reflective structures can be configured into a dual image mirror system (dual Image mirror system), where each light reflection portion is combined with two reflection structures before being measured by a camera. However, the light is combined with more or less than two reflective surfaces before being measured by a camera without departing from the scope of the invention.

The detection member disclosed herein can be further used in other applications. Irradiation of the base of a bottle allows the filling level detection to be observed. Furthermore, the detecting member can be further used for detecting floating and sinking foreign matter.

One embodiment of the invention includes a method of detecting a bottle. A method for detecting a variation in a product in a partially transparent container can include transporting the container for inspection; directing light from a source below the container toward a relevant range of the bottle such that when the reflected light extends at different angles of reflection Light can reflect the relevant range; receive each reflected light to capture two stereoscopic images of the range; and analyze the light patterns of the two solid mirror images to distinguish the variation of the product within the container Different types. A conveyor belt and each support belt move the bottle/container to a position close to or above one of the light sources. The light source can be a laser diode, or an infrared source, or other forms of light source (such as Xenon strobe, Quartz Halogen, laser, visible light, ultraviolet light (UV) ), and infrared light (IR)). A camera is utilized such that a portion of the light reflected by one of the objects in the bottle is detected. It is possible to use each of the electronic sensors instead of using each camera, or each electronic sensor is used in combination with each camera. Light can be transmitted from the base of the bottle towards the neck.

The various embodiments described above utilize a complex distributed computing system (leverage a complex computing system), which includes a human machine interface that can be used to build and construct this detection system, a production line that can be used to track bottles, camera and flash triggers, and exit control. The Line Control Module and a Vision Engine are used to perform the timely computer image calculations required to support the image.

To support real-time detection requirements of more than 1,500 bottles per minute, visual engine components can be implemented as timely, multi-threaded software applications. The timely manner of the system of the present invention is to determine the need for the tested bottle to be classified as a stolen or non-smoked product within a thousandth of a second of the bottle reaching its detection zone. This arbitrary multi-threaded implementation facilitates execution optimization by causing multiple bottle images of the captured bottle to be processed simultaneously. In addition, multi-threaded actions can be used to support pipelining of various processes, including photo capture (acquisition of images) of the next bottle under current bottle handling. While the multi-threaded software infrastructure is an important element of the systems and methods of the present invention described herein, it is critical to simultaneously process the real-time algorithms of the various images. It is worth noting that the contaminant analysis described here is a probabilistic model and moves towards a deterministic model. In the case of a probability mode, the probability is attributed to the object being tested, that is, the distribution of likelihood determines that the target is the probability of the contaminant. This means that an acceptable target (ie, the mold mark on the bottle) may be identified as a contaminant. Furthermore, a pollutant can also be identified as a safe person. Conversely, a deterministic pattern is from a known condition or initial state. Produce the same output. In other words, even with the same bottle input image and the same sensitivity threshold setting on the software, various calculations always produce the same result. The advantage of this system is that after thousands of known off-sweep bottle images are offline sweeps, the software can be trained so that the presence of contaminants leads to positive detection of contaminants. / Identification.

The multi-threaded, deterministic infrastructure of this detection system is closely related to the data associations handled by various threads. These associations can be used to distinguish between lying objects in the bottle and various exterior features, water droplets or dust. In addition, as long as any detection thread finds a stolen item in a bottle and by labeling the bottle as a counterfeit, such association can be used to minimize the operating time of a known bottle for these calculations.

Computer vision algorithms that support this application include a combination of various 2D and 3D calculations, each of which is used to filter each image, including but not limited to grayscale, texture (texture) ), and pixel concentration thresholding. If such low-cost 2D calculations do not absolutely assert the innocence of the bottle, then a wider 3D image classification will be resorted to. The complexity of the 2D calculations used can be configured based on the form of the various bottles and their contents.

The following sequence is the implementer of the possible test calculations:

1. Calibrate the camera with a calibration target to estimate the relative positioning of the camera, mirror, and inspection station.

2. Grab the image from the camera and parse the solid mirror image.

3. In the solid mirror image, a calibration map transformation is applied to both images to obtain a map image. Figure 8.

4. On 2D, use a gap between the aperture of the light exiting the base of the bottle to define the action line 910 AKA "smiley". This smile defines the base inside the bottle. Figure 9.

5. Find the outline of the base of the bottle and define the relevant area (RO1) 1010. Figure 10.

6. Use various 2D segmentation methods to define the edges 1110 of the "smiley face"/action line. Figure 11.

7. The compensation plot is that the line of action is not fully represented in the image.

8. Use 2D filters to smooth-out the line of action.

9. Assign ROIs to each image based on the corrected smiles.

10. Scan the previously defined ROI and use various 2D segmentation techniques for potential artifacts. If no sputum is present, the bottle can be declared a "good" bottle and the calculation can be aborted at this point.

11. According to the ROIs described above, binocular disparity is utilized to create a 3D mode of the bottle.

12. Filter out noise generated by bubbles or small particles in the liquid contents.

13. Use the 3D points thresholding to determine the potential stolen goods in the bottle, and if there are no coordinate points that violate the set threshold, then the bottle The child can be declared as a "good" bottle and the calculation can be aborted at this point.

14. Again, the binocular disparity information on the line of action can be used to define the cylindrical configuration of the bottle, and thus the 3D coordinate points inside and outside of the walls. Figure 12.

15. Location, density, profile, and other threshold attributes can be used to declare if there are any stolen goods in the bottle.

While the contents and advantages of the present invention have been described in detail, it is understood that various changes, substitutions and changes may be made without departing from the spirit and scope of the invention as defined by the appended claims. produce. In addition, the scope of the invention is not intended to be limited to the particular procedures, machinery, manufacture, compositions, compositions, methods, and steps described herein. The person skilled in the art, from the corresponding aspects described herein, will be able to use the present invention to understand the processes, machines, manufactures, compositions, devices, methods, or steps that are present or to be developed. The function or the result of achieving substantial results. Accordingly, the scope of the appended claims is intended to cover all of the processes, machinery, manufacture, compositions, devices, methods, or steps in the scope of the invention.

10‧‧‧ camera

20‧‧‧Mirror

25‧‧‧Mirror

30‧‧‧ Images

35‧‧‧ Images

45‧‧‧Light source

100‧‧‧Detection system or component

200‧‧‧ bottles or bottles

F‧‧‧Shopping

Claims (23)

A system for detecting commercial variations in a transparent or translucent container, comprising: a. a light source that directs light toward a region of interest of the container such that when the reflected light extends differently When the angle is reflected, the light can reflect the relevant range, wherein the light source is located below the container; b. a camera that can receive the reflected light, wherein the parallel reflected light travels to a common part of the light sensors The non-parallel rays travel to different portions of the light sensors; c. a first mirror that reflects the reflected light to the camera; and d. a processor that receives at least the camera Two images, and can distinguish between different types of commodity variations in the container by analyzing the light pattern. The system of claim 1, further comprising a second mirror. The system of claim 1, wherein the light source is collimated light. The system of claim 1, wherein the light source is a polarized light. The system of claim 1, wherein the relevant range is a bottom half or a bottom third of the container. The system of claim 1, wherein the image of the container produced by the first mirror and the second mirror is different from between 2 and 45 degrees. The system of claim 1, further comprising a base above the light source. The system of claim 1, wherein the processing method is deterministic analysis. The system of claim 1, wherein there is a gap between the bottom of the container and the base. The system of claim 1, further comprising another camera. The system of claim 1, wherein more than 100 bottles per minute are detected by the system. The system of claim 1, wherein the base axis is at an angle of about 20 to 25 degrees from the first mirror axis. The system of claim 1, wherein the base axis and the second mirror axis are at an angle of about 10 to 30 degrees. A method for detecting a variation in a product in a portion of a transparent container, comprising: a. transporting the container for inspection; b. directing light from a source below the container toward a relevant range of the container such that when the reflected light extends across the different reflections At an angle, the light can reflect the relevant range; c. accept each reflected light to capture two stereoscopic images of the range; and d. by analyzing the light patterns of the two solid mirror images, Differentiate the different types of commodity variations in the container. The method of claim 14, further comprising transporting the container above a base and having a light source below the base. The method of claim 15, wherein the light source is a collimated light. The method of claim 16, wherein two cameras placed at different angles are used to determine whether there is a foreign substance inside or outside each container. The method of claim 14, wherein the method of analysis is a deterministic analysis. The method of claim 14, wherein the range is the lower half of the container. The method of claim 14, wherein the analysis method is multi-threading. The method of claim 20, wherein the multi-threading method comprises simultaneously analyzing a plurality of solid mirror images, wherein the plurality of solid mirror images comprise the two solid mirror images. The method of claim 14, wherein the analytical method comprises a pipelining process. The method of claim 22, wherein the flow-through operation method comprises simultaneously analyzing a light pattern from the two solid mirror images, and acquiring the solid mirror images of the next container.