Maintaining image integrity within a skinner

Abstract

For a skinner using cameras in an optical system to surveil a hazardous volume beside the blade from both ends, the windows are liable to smearing or blockage, preventing the cameras from seeing the position of the operator's hands. The invention provides software to be run in its internal processor. Each new frame is compared with a pre-loaded training picture. Differences are accumulated over a period of time, on a pixel-by-pixel basis and in real time. If the accumulated differences exceed a pre-arranged number, the processor stops the skinner motor and initiates an appropriate operator cleaning process.

Description

[0001] Title: Maintaining image integrity within a skinner.

[0002] Field:

[0003] The invention relates to improvements to machines known as “skinners” or as “flaying machines”, and to de-rinders.

[0004] The kind of skinner referred to herein is a stand-alone machine, not automated and linked to a conveyor. An operator stands at the machine, grasps an item to be skinned, and pushes the item over a horizontal surface toward the blade, using his or her properly gloved hands. A Ptransverse gripping roller grips the underneath of the item and carries it with power assistance toward a rear, fixed blade. Waste material taken in a thin layer passes under the blade; food material passes over it. For example an ox tongue is skinned in several passes to remove the epithelium, fascia, and other materials covering the muscle. In order to protect the operator, some machines of the type described use optical sensing of an instantaneous position of a worker’s hand or hands in relation to the fixed blade and may halt the machine in order to protect the user’s hands from trauma inadvertently caused by the blade. In particular the invention relates to software within a processor for maintaining the image quality within rays of image-forming light arising from a hazardous volume in proximity to the blade of a skinner and passed to one or more cameras.

[0005] Definitions:

[0006] Hazardous volume.

[0007] The hazardous volume (with respect to an operator’s fingers as he or she advances the item to be skinned) is over the top of the slowly rotatable gripper roller 103 (see Fig 1) and in particular it is just in front of, and includes an edge of, a fixed blade 105 and is situated across the rear of an operator work area, in front of the blade. The hazardous zone, together with a surrounding volume, has an extended length and is viewed by cameras from each end of the blade, unlike the “from above” approach used for bandsaws. During use the gripper roller is rotated and tends to draw articles to be skinned over the fixed knife 105. In the Applicant’s models, the gripper roller surface moves at typically 700 mm per second, under operator control. The stopping time is less than 20 milliseconds. That determines an example fore-and-aft axis for the hazardous volume to be at least (0.7 x 20 mm) = 14 mm depth in thedirection toward the operator, to have a height of about 15 mm above the blade, and to have a length of about 500 mm which is the length of the typical blade.

[0008] Bundle of image-forming rays of light

[0009] This is abbreviated in the following text to “image light”.

[0010] Image integrity

[0011] This refers to any effect that partially or completely prevents an operator’ s glove anywhere within the hazardous volume from being seen by a camera. Effects including physical blocking by opaque items in contact with a viewing window between the hazardous volume and a camera, or translucent smearing of the window by fluids or fat exuded by items to be skinned.

[0012] Camera

[0013] This name refers to cameras used in machine-vision applications, typically involving a constant stream of captured images presented to a processor. Currently preferred Teledyne “Blackfly” cameras are described at https: / / www.teledynevisionsolutions.com / en-150 / products / blackfly-s-usb3 / ?model=BFS-U3-04S2C-CS&vertical=machine%20vision&segment=iis

[0014] Processor

[0015] This refers to a programmed image processor also having software commands that implement the concepts of this invention. The Applicant’ s products include an embedded computer, preferably one capable of running under a high-level language such as WINDOWS® or LINUX®. The embedded computer is capable of image analysis and network communication according to the Applicants’ software, which may be provided in any compatible language. The invention may be enacted in another processor or computer not involved in user hand protection, but it is convenient to use the same one.

[0016] Background:

[0017] The Applicant’s series of imaging optical sensing systems that are adapted to detect an operator's hand close to, and especially in dangerous proximity to a cutting blade of a machine began with WO2017196187A1 for a bandsaw. At least one camera looks down from above on to the bandsaw table and may cause the bandsaw blade to be stopped within 20 milliseconds after a worker’s hand is detected as being dangerously close to the blade.The Applicants have described a skinner or derinder including an optical safety protection apparatus in PCT / NZ2020 / 050142. A skinner has a different optical configuration to that of a bandsaw, because of obscuration by materials above the hazardous volume, so that a downward-looking camera can’t be used. Cameras are directed horizontally along the fixed skinner blade from either side through an apparatus including deflecting mirrors that extend the effective distance between the camera and the viewed object.

[0018] Technical Problem to be solved:

[0019] Optical sensing of a stray hand within the hazardous volume of a manually operated skinner provides operator safety. Complete reliability of any sensing method should be ensured, therefore a reasonably clear view into the hazardous volume beside the entire blade is maintained in order to discover and identify, in real-time, the position of one or more identifiable coloured objects (gloved hands) in relation to the blade inside a hazardous volume, and take action as required.

[0020] Maintaining that clear view requires optical surfaces to be unobscured. Optical surfaces include surfaces of reflective and transmissive optical components, including at least one mirror, at least one window into the working environment, and at least one lens. The surfaces may become obscured over time by, for example, condensation on mirrors, lenses or the windows, a smeared window, and, in particular, immediately by items covering the window at each side of the working area (See Fig 2). The operator may place partly skinned items on the machine surface adjacent one of the windows. One estimate for an unaware operator is that as many as 1 in 10 skinning operations are blocked by such items so that the hazardous volume is not properly supervised by the cameras. This invention relates to use of software to solve the technical problem by continuously monitoring views of the hazardous volume.

[0021] Object

[0022] A first object of the invention is to provide an image quality maintenance function for use in in a skinner machine since the camera or cameras of the skinner work to recognise an operator’s glove and its position in relation to the fixed blade. A related object is to provide a safe working environment for an operator of a skinner by detection of poor image quality and requesting the operator, through an interface, to clean accessible surfaces. A further object is to at least to provide the public with an alternative choice of skinner.Summary of Invention:

[0023] In a first broad aspect, the invention provides a skinner machine for use by an operator; the skinner machine having a horizontal working surface having a near (that is, near to the operator) edge, a left side border and a right side border and including a power-driven gripping roller intended to carry an item to be skinned against a blade of a fixed knife located along a far edge of the working surface; the skinner machine having an optical safety system intended for detection of a distinctive object, namely a part of a gloved hand of an operator, within a predetermined, illuminated hazardous volume located along the edge of the blade and toward the position of the operator; the optical operator safety system including two or more cameras each viewing the hazardous volume along an optical image path from each end of the hazardous volume; the optical image path to be monitored including the hazardous volume and an optical window exposed to the working surface and separating the working space from a sealed space in front of a lens of each camera; and the output of said one or more cameras is sent to an image processor within an embedded controller wherein the embedded controller receives, and is programmed to monitor images collected by the cameras during use, to recognise indications of loss of image integrity, and to initiate remedial action if image integrity is diminished.

[0024] Preferably the embedded controller is programmed in order to detect a static physical obstruction within any optical image path, determined to exist if the image seen by any one or more of the electronic cameras, in a region over or near an image of the hazardous volume (herein the “active image area”) includes a portion that is static over a period of time; the controller being capable of maintaining, for each camera, an array in memory representing current pixel values within the active image area that are capable, when in use, of being refreshed for each frame, a reference image or training picture, and a non-refreshed event counter for each pixel within the active image area.

[0025] Preferably the embedded controller also maintains a previously collected training picture of the active image area; one for each camera, as pixels; each pixel of the training picture being accompanied by a variance or threshold value memory value derived from the process of collection of the training picture from a number of frames.

[0026] Preferably if a difference exceeding the variance is found as the result of any single pixel comparison against the training picture, the non-refreshed event counter associated with that pixel is incremented.Preferably, after a period of time has elapsed, every pixel counter that has exceeded a predetermined number and is contiguous with other pixel counters that have exceeded the predetermined number is taken to comprise a group of pixels defining an image of the static physical obstruction and a sequence to have the obstruction removed is entered; the sequence including the step of resetting the non-refreshed event counters, and commencing one or more further actions selected from the range of: bringing the skinner motor that drives the gripper roller to a slow halt, sending a sound alarm to the operator, sending a visual alarm to the operator, sending a message to the operator’s console, and sending a message over an intranet to a supervisor.

[0027] Optionally the skinner machine as previously described in this section uses the embedded controller in order to detect a diffuse obstruction or smear within the optical image path; detection including the range of: calculating image contrast by comparison on a pixel-by-pixel basis with the training picture, and calculating image contrast by comparing an intensity of identified highlights with an intensity of identified lowlights as compared to that of the training picture, and in event of a diffuse obstruction or smear being detected, a sequence to have the obstruction or smear removed is entered.

[0028] Alternatively, the embedded controller is programmed in order to detect an existence of a reduction in image contrast over time resulting from contamination, of any optical surface within the light path by presence of a material selected from a range including a solid or liquid arising from the material being skinned, a smear, or a condensed liquid or contamination in an image received by any camera; existence may be determined by computing a distribution or histogram of pixel brightness over the image and if the mean high level and the mean low level approach each other, the processor will prevent operation of the skinner motor until the operator has had the optical surfaces cleaned.

[0029] Optionally, a test object comprises the illuminated LEDs behind the window on the opposite side of the skinner work surface, and an image of a plurality of the test objects is evaluated for contrast and presence of black levels before the operator starts to use the machine.

[0030] Alternatively, the test object comprises a clean operator’s coloured glove brought into the hazardous volume of the skinner machine while the gripping roller is inactivated, and the image of the glove is evaluated for brightness and contrast in comparison with stored data before the operator starts to use the machine.Description:

[0031] The descriptions of the invention provided herein are given purely by way of example and are not to be taken as in any way limiting the scope or extent of the invention. Although some specific numbers have been established by trial, they are not limiting. Throughout this specification, the word “comprise” and variations such as “comprises” or “comprising” should be understood to imply the inclusion of a stated option, integer or step, but not the exclusion of any other option, integer or step. Reference to cited material or to information in the text should not be understood as a concession that the material or information was part of the Common General Knowledge or was known in New Zealand or in any other country. LIST OF FIGURES

[0032] Fig 1: Diagrammatic vertical cross section of a skinner including some sites where degradation of images may occur.

[0033] Fig 2: Oblique view of one side of a skinner’s work area from the operator’s side, showing the lights, the camera window, and an obstruction.

[0034] Introduction:

[0035] The technical function of the software to be described is operator protection. The software runs in an embedded processor, usually in addition to software used for optical identification of a hazard ( a gloved finger too close to a blade) within a skinner machine. That service requires continuous and effective surveillance of an elongated hazardous volume, as described in “Definitions”, above. The operator must wear a coloured glove; blue or green, which is a usual requirement for hygiene purposes.

[0036] According to this invention, the hazardous volume is surveilled from both ends and preferably with two cameras at each end which are used to provide separated viewpoints.

[0037] PREFERRED EMBODIMENTS

[0038] All improvements described herein have been made within the software of the embedded processor.

[0039] At the time of filing, preferred frame rates for each of the four cameras is 300 per second. The preferred processor can perform the calculations to be described in between frames. In the current embodiments, the images to be analysed (as described below) sampled from the full frames are in sub-frames about 270 x 170 pixels in size, in colour. Data from thecameras is preferably digital. It may be useful to retrieve only the sub-frames from the cameras, if possible.

[0040] Fig 1 is a diagram of a vertical section through a skinner 100, including optical apparatus and the hazardous volume 101. There is a left side optical tunnel, 102’ and a right side optical tunnel 102. The optical tunnels are one optical solution to the problem of transferring a relatively proximity-independent view from within the hazardous volume 101 by distancing the lens from the volume to be surveilled. Other solutions may not require tunnels having optical lengths. (Note that the skinner is bilaterally symmetrical, for views of the surveilled volume, including the hazardous volume from both ends). The volume under surveillance contains an elongated, illuminated hazardous volume 101. Beams of image-forming rays of light (image light) may arise from operator’s gloves and emerge from the hazardous volume to the left and to the right, toward eithera window 107, 107’ along a channel made during the process of skinning an item against a blade. Appropriate illumination is preferably delivered into the hazardous volume from arrays of preferably white light-emitting diodes or LEDs 104 located around the viewing window near each end of the blade (Fig 2). Image light from a glove passes through a protective window 107, 107’ into a tunnel, is deflected off an inclined front-silvered mirror 103, 103’ and passes through the tunnel along optical axis 108, 108’ to reach a lens 104, 104’ of an camera 105, 105’. Although the drawing does not show it, at the time of filing, there are two such cameras at the end of each tunnel, for a total of four cameras and four series of frames to be tested in “real-time”.

[0041] Data passes along lines 110 (right) and 111 (left) to the embedded processor 112. Outputs from the processor 112 through control line 113 may comprise either slow or emergency-fast stopping operation of the skinner motor 114 that drives the gripping roller. Further information is displayed to the operator, to a supervisor, or to a maintenance engineer who may be remote on a screen 116 or a link. The view seen by any camera can be displayed. The operator may also be alarmed by lights, horns, bells, buzzers and other acoustic signals (not shown), and he or she will be aware of the motor 114 ceasing to operate either as a consequence of hand proximity to the blade, or because of a detected obstruction to any beam of image light.

[0042] The image path of an optically sensitive skinner may be adversely affected if physical objects 202 (See Fig 2) obstruct either beam of image light. Fig 2 is an oblique view of window 107 from within the skinner. The other side is a mirror image of Fig 2. The operator is placed farto the left and engages materials to be skinned on the motorised gripping roller 106 which brings the materials on to an edge of the blade 203. The window is extended as 201 over a number of white LEDs 104 that provide light into the hazardous volume from each end. The example object 202 is a piece of flesh to be recognised as a static physical obstruction.

[0043] EXAMPLE 1: Image Degradation caused by a Static Physical Obstruction:

[0044] It is always undesirable for the optical protection apparatus to be disabled because the operator is then at risk, but the operator of apparatus lacking the invention may not be aware of degraded optical performance. Sometimes an operator will leave a piece of meat 105, 202 on the machine to one side of the blade and gripping roller, obstructing the viewing window. (See Fig 2) The operator may be unaware that the cameras can’t clearly see his or her gloved hands.

[0045] The image processor of the skinner may be programmed in any suitable language to look for signs of obstruction and take remedial action according to the steps below. The embodiment below illustrates how image data may indicate that a window has been at least partially covered by a static item for more than an allowed predetermined time.

[0046] The processor will preferably include steps to evaluate each image for degradation before the operator starts to use the machine, or during each shift, and to evaluate each beam of image light for sudden onset of degradation (including presence of an obstructing object), at all times during operation of the skinner.

[0047] Existence of an obstruction 202 especially outside a window, between the knife area and an interior of a sealed tunnel may be established by the processor from video information in incoming video lines 110 and 111 of Fig 1. The processor is empowered to collect a train of images over a period of time and compare each one with the training picture, to establish if at least a part of the image is static over the period of time but the gripper roller is turning. The raw image seen by any camera is dissected as follows: An outer “warning zone” having an oval upper boundary is declared over the upper surface of the blade. An inner ’’Hazardous zone” is set up inside that warning zone. The warning zone dimensions are about 270 pixels across, and 170 pixels high and pixels are handled in colour. The algorithm used by this invention concentrates on the warning zone and the hazardous zone, and if a drop in visual clarity should occur, of stopping the gripping roller.

[0048] This is a list of algorithm process steps, at least in principle.1. Image checking is restricted to an “active image area” about 270 pixels across, and 170 pixels high, which is handled in colour. That is true for the configuration in the prototypes. Any image pixel within, or a memory cell holding a training picture value such as R, G, B and deviation, a fresh image representing it as R, G, B, together with an associated event counter is easily accessible by conventional addressing.

[0049] Alternatives to pure R, G and B values may be used. They include colour and brightness.

[0050] 2. Each camera of each skinner is set up with a Training Picture of the image that it sees, when manufacture is completed. For each pixel, an averaged R, G and B colour, and deviation (including noise) is recorded from perhaps 100 frames for later use. The process may be repeated as a field modification.

[0051] 3. During use, the obstruction-detecting software will notice contiguous blocks of divergent pixels, so takes action if blocks of 20 or more contiguous pixels are identified as different. (20 is a convenient number and may be varied).

[0052] 4. Every pixel in the active image area of each frame is tested separately against the corresponding pixel in the Training Picture. If there is a difference beyond the recorded variance for that pixel, the event counter for that pixel is incremented by 1. Then, for the next frame, the same test is made and the counter may be incremented again. After for example ten seconds, data from 3000 frames have been collected for evaluation. Each event counter may have reached a counter threshold of 3000 if the difference between it and the pixel in the Training Picture exceeds the variance.

[0053] Photon noise and electronic noise may exist, since the images are not brightly lit.. 5. The processor may use a “sum of event counters” as a simple test before entering the next stage, when ten second has elapsed.

[0054] 6. Isolated event counters which hold high counts are ignored so that only contiguous high counts from blocks of for instance 20 contiguous pixels are taken as significant. Testing for whether one high count is contiguous to another involves considering the adjacency of the addresses of the two. On the other hand, counts taken from moving images (as will occur during normal operation) are more likely to be wiped across the view and will not reach the counter threshold for any one pixel.

[0055] 7. After applying the test for contiguousness, the processor may decide that enough counters for contiguous pixels hold high counts as shown by a high Total Count, and command the motor of the skinner to stop. The processor has found a part of theimage from that camera that is (a) unlike the Training Picture and (b) appears to be static. The processor may:

[0056] a. Bring the skinner motor that causes the gripper roller to turn to a slow halt. b. Send a message to the operator, in one or more ways including: Aural warning, flashing light warning, and a displayed message on the operator’s console 116. The message may identify the location of the obstruction, at least by left or right side and invite remediation. The processor can show the operator a camera image of the suspected obstruction so that the operator can remove it.

[0057] c. Send a message with an image over an intranet to a supervisor.

[0058] 8. Event counters are reset to zero after the 10 second evaluation or during a return to operation. The ten second period is a somewhat arbitrary “operator behaviour” factor. The count may be established over 5 or 3 seconds for example, or a rolling reset always representing the “last ten seconds” that is updated every second may be maintained.

[0059] Behavioural conditioning is in effect. If the operator experiences machine shutdown after a particular kind of behaviour, he or she may avoid placing items in front of the windows even without being aware of having undergone a training step.

[0060] EXAMPLE 2: Image Degradation caused by a Smear:

[0061] If the image is obscured, such as by blood, fat or exudate spread on to the window, the cameras may be prevented from clearly detecting blue gloves in a process as described above, detecting a smear in the same way, because there will be a stationary, regional difference as in Step 4 of the list.. Data from pixels representing the smeared area will differ from pixels within the Training Picture. Incidentally, it has been noted that blue light tends to penetrate typical smears better than white light. A smear problem may be apparent in a pre-operation test even before the gripper roller starts to turn.

[0062] The list of algorithm process steps given above may also disclose this kind of obscuration If deposits, condensation, and smears including solids and liquids, on the window exterior appear to exist, the operator can be advised to clean the window or both windows.

[0063] EXAMPLE 3: Slow degradation of the image:Further tests may be applied, for detection of long-term deterioration of the image. Most tests require only software, although one test for window clarity would involve installing lamps for obliquely lighting the window itself with nearby lamps to scatter light off contamination. Image quality may gradually worsen over time, which places the skinner operator at risk. Examples include:

[0064] 1. Deposits, condensation, and smears including solids and liquids, on the window 107.

[0065] The interiors of the tunnels are sealed against ingress of any sort, but seals may break down. Unexpected vapours, such as from evaporating plasticisers, may drift on to optical surfaces over time. Condensation may inadvertently occur because the skinner is operated in a chilled room.

[0066] 2. Condensation on, or deterioration of the front- silvered mirrors 103.

[0067] 3. Condensation or deposits on or in a camera lens 105. In some configuration, solid particles may fall on to the lens or other upward-facing optical surfaces.

[0068] 4. Condensation or deposits on the sensor chip of the camera 105.

[0069] 5. Scattered light and stray light.

[0070] 6. Light levels -related to illumination efficacy - can be checked.

[0071] 7. Image shifting, perhaps caused by vibration from the stopping motor affecting camera mounts or mirror supports will become apparent by comparison with the corresponding Training Picture.

[0072] 8. Intrinsic faults within the image sensor chip of the camera, or in related circuitry. A supervisor may decide to reset the Training Picture which would cancel out any record of previous loss of image integrity.

[0073] Algorithms relating to slow degradation have not yet been put into practice but would include the following aspects:

[0074] Large droplets of condensation would be noted by the algorithm of Example 1, but would be rejected under the test for “contiguous changed pixels”. They are most likely if the temperature of the working environment changed. The operator would see external condensation. Internal (inside the tunnel) condensation may be avoided by including a container of a drying agent such as silica gel inside each tunnel at the time of manufacture. Condensation either in small-droplet form or as solid material on any optical surface is likely to scatter bright light over black parts of an image so that backgrounds become brighter, atleast in proximity to the image of the light source. Since the cameras on each side of a skinner see the LEDs of the light source on the opposite side, a target is readily available. Accordingly, an algorithm looking for small droplets or solid coatings over an entire optical surface may construct a densitometer trace (in memory) from the middle of any one lit LED to the middle of the next and examine it for a failure to approach a black level. Any such coating would tend to scatter light.

[0075] Inside a camera, the CCD or other electronic parts may develop a defect. Well-known CCD defects include dark or bright pixels and lines.

[0076] An example of a test target is a clean coloured glove, worn by the operator and brought into the hazardous area before the roller has started to turn.

[0077] The contrast range of an image of a known object such as a LED lamp can be determined with reference to the black level surrounding each lamp image. Desirably, the processor, which is relatively powerful, would display the outcome of a diagnostic procedure and at least would specify whether the left or the right side tunnel is affected. It may display a visual image as collected on the operator’s LCD screen or similar console.

[0078] Evaluation for brightness is capable of sensing illumination quality and camera function. Optionally, evaluation is repeated whenever the gripper roller is not in use.

[0079] ADVANTAGES

[0080] A skinner that monitors image quality as described in this document is less likely to fail to see a part of the operator’s hand in a dangerous proximity to the skinner blade, within the hazardous volume.

[0081] The improvements are made in software.

[0082] Finally it will be understood that the scope of this invention as described and / or illustrated herein is not limited to the specified embodiments. Those of skill will appreciate that various modifications, additions, known equivalents, and substitutions are possible without departing from the scope of the invention as set forth in the following claims.

Claims

1. WE CLAIM1. A skinner machine for use by an operator; the skinner machine having a horizontal working surface having a near (that is, near to the operator) edge, a left side border and a right side border and including a power-driven gripping roller intended to carry an item to be skinned against a blade of a fixed knife located along a far edge of the working surface; the skinner machine having an optical safety system intended for detection of a distinctive object, namely a part of a gloved hand of an operator, within a predetermined, illuminated hazardous volume located along the cutting edge of the blade and toward the position of the operator; the optical operator safety system including two or more electronic cameras each viewing the hazardous volume along an optical image path from each end of the hazardous volume; the optical image path to be monitored including the hazardous volume and an optical window exposed to the working surface and separating the working space from a sealed space in front of a lens of each camera; characterized in that the embedded controller receives, and is programmed to monitor images collected by the cameras during use in order to recognise one or more indications of reduction of image integrity and to initiate remedial action if image integrity is reduced.

2. The skinner machine as claimed in claim 1, characterized in that the embedded controller is programmed in order to detect a static physical obstruction within any optical image path, determined to exist if the image seen by any one or more of the electronic cameras, in a region over or near an image of the hazardous volume (herein the “active image area”) includes a portion that is static over a period of time; the controller being capable of maintaining, for each camera, an array in memory representing current pixel values within the active image area that are capable, when in use, of being refreshed for each frame, a reference image or training picture, and a non-refreshed event counter for each pixel within the active image area.

3. The skinner machine as claimed in claim 2, characterized in that the each pixel of the training picture is accompanied by a variance or threshold value derived from a historic process of collection of the training picture from a number of frames.

4. The skinner machine as claimed in claim 3, characterized in that if a difference exceeding the variance is found as the result of any single pixel comparison against the training picture, the non-refreshed event counter associated with that pixel is incremented.

5. The skinner machine as claimed in claim 4, characterized in that after a period of time has elapsed, every non-refreshed event counter that has exceeded a predetermined number and is contiguous with other event counters that have exceeded the predetermined number is taken to comprise a group of pixels defining an image of the static physical obstruction and a sequence to have the obstruction removed is entered; including the step of resetting the non-refreshed event counters, and commencing one or more further actions selected from the range of: bringing the skinner motor that drives the gripper roller to a slow halt, sending a sound alarm to the operator, sending a visual alarm to the operator, sending a message to the operator’s console, and sending a message over an intranet to a supervisor.

6. The skinner machine as claimed in claim 1, characterized in that the embedded controller is programmed in order to detect an existence of a reduction in image contrast over time resulting from contamination, of any optical surface within the light path by a material selected from a range including a solid or liquid arising from the material being skinned, a smear, or a condensed liquid or contamination in an image received by any camera, may be determined by computing a distribution or histogram of pixel brightness over the image and if the mean high level and the mean low level approach each other, the processor will prevent operation of the skinner motor until the operator has had the optical surfaces cleaned.

7. A skinner machine as claimed in claim 1, characterized in that a test object comprises the illuminated LEDs behind a window on the opposite side of the skinner work surface, and an image of a plurality of the test objects is evaluated for contrast and presence of black levels before the operator starts to use the machine.

8. A skinner machine as claimed in claim 7, characterized in that the test object instead comprises a clean operator’s glove brought into the hazardous volume of the skinner machine while the gripping roller is inactivated, and the image of the glove is evaluated for brightness and contrast in comparison with stored data before the operator starts to use the machine.

9. The skinner machine as claimed in claim 1, characterized in that the embedded controller is programmed in order to detect a diffuse obstruction or smear within the optical image path; detection including the range of: calculating image contrast by comparison on a pixel-by-pixel basis with the training picture, and calculating image contrast by comparing an intensity of identified highlights with an intensity of identified lowlights as compared to that of the training picture, and in event of a diffuse obstruction or smear being detected, a sequence to have the obstruction or smear removed is entered.

Images