WO2013053597A1

WO2013053597A1 - Method and system for detecting a louse on fish

Info

Publication number: WO2013053597A1
Application number: PCT/EP2012/068921
Authority: WO
Inventors: Baard RØSVIK
Original assignee: SALVISION AS
Current assignee: SALVISION AS
Priority date: 2011-10-12
Filing date: 2012-09-26
Publication date: 2013-04-18
Anticipated expiration: 2014-04-12
Also published as: NO333499B1; NO20111385A1

Abstract

The present invention relates to a method for detecting a louse on a fish, comprising the steps of receiving an image of a fish; searching for pixels having colors within a background color interval and setting those pixels to a default pixel value; searching for pixels having colors within a fish surface color interval and setting those pixels to the default pixel value; and searching for a predetermined louse-shaped object in those parts of the image having a pixel value different from the default value. Then, the method comprises the step of determining if the predetermined louse-shaped object is found in the image. If it is found, the method is performing the steps of: providing a sub-image of the image, where the sub-image is containing the area of the image in which the predetermined louse-shaped object is found; and storing information about the sub-image, where the storing of information comprises storing information about the position of the sub-image in the image.

Description

METHOD AND SYSTEM FOR DETECTING A LOUSE ON FISH

FIELD OF THE INVENTION

The present invention relates to a method and system for detecting a louse on a fish.

BACKGROUND OF THE INVENTION In the fish farming industry, lice have been and still are a huge problem. For example is salmon (especially Salmo Salar) in fish farming facilities vulnerable to the salmon louse (Lepeophtheirus salmonis). Also other fish species, such as rainbow trout (Oncorhynchus mykiss), cod species etc vulnerable to different types of lice. There has been several attempts on both monitoring the number of lice in the fish farming facility and on removing the lice from the fish in the fish farming facility.

The salmon louse has a life cycle of ten life cycle stages, where stages four to seven are the chalimus stages where the louse is attached to the salmon and eats from the salmon, and where stages eight to ten are the adult stages where the louse is mobile. In these last stages, it is also possible to differentiate between male and female lice (see for example http://en.wikipedia.org/wiki/Salmon_louse). There are other types of fish lice that have similar life cycles and similar appearance.

Today, the monitoring of the number of lice in a fish farming facility, such as a fish farming tank, fish farming cage etc, is performed manually, by catching one or a number of fish out from the facility and then count the number of lice on each fish manually. Alternatively, an image recorder (film or individual images) can be submerged into the facility and the number of lice can be counted manually by viewing the image recordings.

It should be noted that the total number of lice is at considerable level if there is an average of 0.5 louse per fish. At this level, delousing must be performed.

Consequently, it is a technical challenge to provide a method which is capable of detecting the one louse on a fish.

The object of the invention is to provide a method and system for detecting the presence of a louse on a fish. Moreover, it is an object to count the number of lice if more than one louse is present. It is also an object of the invention to provide a method and system where it is possible to detect the stage in which each louse on the fish belongs to. It is also an important object of the invention to provide a method and system which is fast and efficient.

SUMMARY OF THE INVENTION The present invention relates to a method for detecting a louse on a fish, comprising the steps of:

- receiving an image of a fish;

- searching for pixels having colors within a background color interval and setting those pixels to a default pixel value;

- searching for pixels having colors within a fish surface color interval and setting those pixels to the default pixel value;

- searching for a predetermined louse-shaped object in those parts of the image having a pixel value different from the default value; - determining if the predetermined louse-shaped object is found in the image, and if it is found, performing the steps of:

- providing a sub-image of the image, where the sub-image is containing the area of the image in which the predetermined louse-shaped object is found; and

- storing information about the sub-image, where the storing of information comprises storing information about the position of the sub-image in the image.

In one aspect, the predetermined louse-shaped object comprises a first, substantially circular object, a second, substantially circular object connected to the first object and a third, substantially rectangular object connected to the second object.

In one aspect, the louse-shaped object comprises an image of a louse. In one aspect, the image of the louse is retrieved from a database of louse images.

In one aspect, the step of storing information about the sub-image further comprises:

- storing the width and length of the sub-image.

- storing the contour of the louse in the sub-image.

In one aspect, the method further comprises the step of:

- determining a life cycle stage of the louse based on the width and length of the sub-image. In one aspect, the received image of the fish is taken under controlled light conditions. In one aspect, the method further comprises the step of performing an image differentiating algorithm before the step of searching for the predetermined louse pattern.

In one aspect, the method further comprises the step of increasing a lice counter for each louse found in the image.

In one aspect, the method is further comprising the step of computing the average number of lice per fish.

The invention also relates to a system for detecting a louse on a fish, comprising:

- an image recorder for recording an image of a fish; - a computer memory;

- a central processing unit connected to the image recorder and to the computer memory, where the central processing unit is configured to perform the method according to the above.

The present invention also relates to a method for detecting a louse on a fish, comprising the steps of:

- receiving an image of a fish;

- searching for a predetermined louse-shaped object in those parts of the image having a pixel value different from the default value;

- determining if the predetermined louse-shaped object is found in the image, and if it is found, performing the steps of: - increasing a lice counter for each louse found in the image.

In one aspect, the method is further comprising:

- computing the average number of lice per fish.

DETAILED DESCRIPTION Embodiments of the invention will now be described in detail with reference to the enclosed drawings, where:

Fig. 1 illustrates the system for detecting a louse on a fish;

Fig. 2 illustrates the steps of a method for detecting a louse on a fish; Fig. 3 illustrates a predetermined louse-shaped object;

Fig. 4 illustrates a screen image of the system running on a computer device, where it is shown that a louse is detected on the fish;

Fig. 5 shows an image as received from the image recorder, where a louse is indicated by a dashed circle;

Fig. 6a shows the image in fig. 5 after the color filtration process;

Fig. 6b shows an enlarged view of the louse found in fig. 6a;

Fig. 7a shows the image in fig. 5 and fig. 6a after a pixel recoding process;

Fig. 7b shows an enlarged view of the louse found in fig. 7a

It is now referred to fig. 1 , illustrating a system 1 for detecting a louse on a fish.

The system 1 comprises an image recorder 2 for recording an image of the fish. The image recorder may record still images (photos) or film, in the present embodiment the image recorder records a still image of the side of the fish.

It should be noted that the system may comprise several image recorders, for example one image of each side surface of the fish. Alternatively, yet another image recorders may be provided to record a top image of the fish, a bottom image of the fish etc.

It should also be noted that the image recorder could record images of only parts of the fish instead of images of the entire fish, i.e. in order to analyze one fish, the image recorder must record several images.

In the present embodiment, the image recorder is a digital camera.

The image recorder 2 is connected to a computer device 3, comprising a central processing unit or CPU 4 and a computer memory 5. The computer device 3 can any type of computer device and is considered known for a skilled person. The computer device 3 is further connected to a user interface, such as a screen, a keyboard, a mouse etc. However, it is also possible to provide the system 1 in one unit, where the unit comprises the image recorder, the computer memory and the CPU in one unit, and where the output from such a unit is a number display showing the average number of lice per fish.

In an industrial setting, the system 1 can be provided with an entrance for the fish and an exit for the fish, in order to ensure that only one fish passes the image recorder simultaneously. Moreover, sensors can be provided in order to be able to record the image of the fish when the fish is in the correct position in front of the camera. Moreover, the image of the fish may be provided when the fish is fetched from water or when the fish is submerged in water.

A method for detecting a louse on a fish is implemented as a computer program running on the CPU 4. Hence, the central processing unit 4 of the computer device 3 is configured to perform the method described below.

In a first step 10 the computer device 3 is receiving an image of a fish from the image recorder 2. The image is provided in an electronic file format, typically a JPG format, a PNG format or another suitable format. An example of such an image is shown in fig. 5, where a dashed circle is indicating the position of a louse. The image in fig. 5 has a size of ca 2.6 MB.

Then, the computer device 3 is manipulating the image by performing the following steps:

In step 12, the computer device 3 is searching for pixels having colors within a background color interval and setting those pixels to a default pixel value. The background color interval is a predetermined interval of colors which are considered to define the background of the image. Hence, if no fish is present, the image will contain only colors within the background color interval.

It should be noted that the term color here is used both for color images, where the pixel may have a value for each of the colors red, green and blue, alternatively a value for each of the colors cyan, magenta, yellow and black. However, the term color may also represent shades of grey in a black and white image.

The default pixel value could typically be a non-color value, for example -1 or pure black or pure white. To obtain this, the image recorder may be directed towards a surface having a color different from the color of the fish and the color of the fish surface.

In step 14 the computer device 3 is searching for pixels having colors within a fish surface color interval and setting those pixels to the default pixel value. The fish surface color interval is a predetermined interval of colors which are considered to define the fish surface of the image. It should be noted that step 12 and 14 may be performed simultaneously. Hence, the method is iterating through the pixels of the image only once while searching for pixels having colors within a background color interval and while searching for pixels having colors within a fish surface color interval. Fig. 6a is showing the image in fig. 5 after step 12 and 14 has been performed.

Here, the default pixel value has been set to white. An enlarged view of the louse as apparent from fig. 6a is shown in fig. 6b.

Hence, since the background and the fish surface will normally form the largest part of the image, searching steps 12 and 14 will have reduced the amount of

information in the image considerably. Typically, only a very small portion of the pixels in the image now do not have the default value.

It should be noted that this will depend on the image, whether or not the image shows the entire fish (i.e. a large portion of the image will contain background colors) or only a section of the fish (i.e. the image will contain no background colors or only a small area with background colors).

The image in fig. 5 has a size of ca 9 kB, i.e. the information in this image has been reduced considerably.

In step 16, the computer device 3 is searching for a predetermined louse-shaped object in those parts of the image having a pixel value different from the default value.

One example of such a predetermined louse-shaped object is illustrated in fig. 3. The predetermined louse-shaped object may be an object or specification making it possible to determine whether a louse actually is present in those parts of the image having a pixel value different from the default value or not. The predetermined louse-shaped object 30 may comprise a predetermined shape as in fig. 3, comprising a first, substantially circular object 31 , a second, substantially circular object 32 connected to the first object 31 and a third, substantially rectangular object 33 connected to the second object 32. A threshold value may be used to determine how close to the image these objects must be in order to define a louse. One additional criterion may be that these objects may be partially overlapping.

Moreover, an additional criterion may be that these objects are oriented along a line I-I.

In an alternative embodiment, the predetermined louse-shaped object comprises an image of a louse. Hence, an image recognition algorithm is performed for comparing the image of a louse and those parts of the image having a pixel value different from the default value. It is possible to provide the system with a database of louse images and retrieve the image of the louse from the database of louse images. The database could comprise several images and compare those parts of the image having a pixel value different from the default value with one or several of the images in the database.

In fig. 6a the louse was found by searching for a predetermined louse-shaped object in the form of a predetermined shape, as the one shown and described above with respect to fig. 3. Yet another test was performed, and the louse was also found by comparing fig. 6a with images of lice.

In step 18 the computer device 3 is determining if the predetermined louse-shaped object is found in the image. Hence the computer device 3 is determining whether or not one or several lice are present in the image.

If the predetermined louse-shaped object is found, the computer device 3 is performing the steps of providing a sub-image of the image, where the sub-image is containing the area of the image in which the predetermined louse-shaped object is found (step 20). Moreover, the computer device 3 is storing information about the sub-image, where the storing of information comprises storing information about the position of the sub-image in the image (step 22). Then, in step 24, the method ends.

The sub-image is shown by a dashed square in fig. 6b, where also the length L and the width W are indicated. If no predetermined louse-shaped object is found, the method starts over again by receiving a new image in step 10. Alternatively, for example if there are no further images, the method ends.

The step 22 of storing information about the sub-image may further comprise storing the width and length of the sub-image. The computer device 3 may determine a life cycle stage of the louse based on the width and length of the sub- image.

The step 22 of storing information about the sub-image may further comprise storing the contour of the louse in the sub-image. The contour of the louse may be stored in the database and may be used in the searching for a predetermined louse- shaped object, step 16 in other images.

In order to improve the method above, the received image of the fish may be taken under controlled light conditions. Hence, the method is assumed to be more reliable with respect to the searching for pixels having colors within a background color interval, the searching for pixels having colors within a fish surface color interval and the searching for a predetermined louse-shaped object in those parts of the image having a pixel value different from the default value. It should also be noted that the method further may comprise the step of performing an image differentiating algorithm after the searching steps 12 and 14, but before step 16 of searching for the predetermined louse pattern. Such an image

differentiating algorithm may be an algorithm where pixel values are recoded, for example to a binary image, for faster shape recognition. The image differentiating algorithm can be performed in several iterations.

Fig. 7a shows the image of fig. 6a where such an image differentiating algorithm has been performed. Here, all pixels in the image are either black or white. An enlarged view of the louse as apparent from fig. 7a is shown in fig. 7b. In fig. 7a the louse was found by searching for a predetermined louse-shaped object in the form of a predetermined shape, as the one shown and described above with respect to fig. 3.

It is also possible to perform pre-processing of the image before step 12 and 14.

It is now referred to fig. 4, showing the user interface 40 of a computer program executed by the computer device 3. It should be noted that this user interface is used for testing of the prototype of the method, and hence there are several possibilities for adjusting different threshold values in order to perform the image differentiating algorithm(s). In addition, the user interface is showing a lot of information about the image to the user, which is strictly not necessary in order to detect whether or not a louse is present on the fish or to count the number of lice on a fish. It should therefore be noted that the method and system according to the present invention does not need a user interface like the one shown in fig. 4.

In fig. 4, it is shown that the user interface 40 shows the image 41 of the fish.

Moreover, the user interface 40 contains a text box 42 where information is shown if a louse is detected. In fig. 4 the information contains the following information:

Lice location = {X=1895, Y=1290}

Lice height = 85

Lice length = 55

Consequently, a louse has been detected, the upper right corner of the sub-image of the louse is in the position (X=1895 pixels (from the right side of the image),

Y=1290 pixels (from the upper side of the image)) of the image, the height of the sub-image is 85 pixels and the length of the sub-image is 55 pixels.

The position of the louse is indicated by a dashed circle 43.

In the present embodiment, a color filter is used when performing step 12 and step 14. The color filter typically defines a color interval with a minimum value and a maximum value for a color value. The maximum and minimum values can be defined in any color system, such as RGB (selecting values for variables Red, Green, Blue), CMYK (selecting values for variables Cyan, Magenta, Yellow, Black), HSL (selecting values for variables Hue, Saturation, Light) or other color system. The color interval can be defined as a minimum value and a maximum value for each variable in the color system or for only one variable in the color system. In addition, a minimum value and a maximum value can be defined for additional properties such as opacity/transparency, reflexivity, etc. The type of color filter and type of color system are considered known for a skilled person and will not be described here in detail.

In order to implement a color filter in step 12 and 14 there are libraries available in many programming languages. Moreover, shape recognition algorithms, alternatively image comparison algorithms, used in step 16, are considered known for a skilled person and will not be described here in detail. Libraries are available for implementing these algorithms in many programming languages as well.

In addition, the method may comprise the step of increasing a lice counter for each louse found in the image. Moreover, the method may comprise the step of counting the number of fish being taken images of, and hence perform the step of computing the average number of lice per fish.

According to the invention a fast and efficient method and system for detecting a louse on a fish is achieved.

Claims

1. Method for detecting a louse on a fish, comprising the steps of:

- receiving an image of a fish;

- determining if the predetermined louse-shaped object is found in the image, and if it is found, performing the steps of:

2. Method according to claim 1, where the predetermined louse-shaped object comprises a first, substantially circular object, a second, substantially circular object connected to the first object and a third, substantially rectangular object connected to the second object.

3. Method according to claim 1, where the louse-shaped object comprises an image of a louse.

4. Method according to claim 2, where the image of the louse is retrieved from a database of louse images.

5. Method according to any one of the above claims, where the step of storing information about the sub-image further comprises:

- storing the width and length of the sub-image.

6. Method according to any one of the above claims, where the step of storing information about the sub-image further comprises:

- storing the contour of the louse in the sub-image.

7. Method according to claim 5, where the method further comprises the step of:

- determining a life cycle stage of the louse based on the width and length of the sub-image.

8. Method according to any one of the above claims, where the received image of the fish is taken under controlled light conditions.

9. Method according to any one of the above claims, where the method further comprises the step of performing an image differentiating algorithm before the step of searching for the predetermined louse pattern.

10. System for detecting a louse on a fish, comprising:

- an image recorder for recording an image of a fish;

- a computer memory;

- a central processing unit connected to the image recorder and to the computer memory, where the central processing unit is configured to perform the method according to any one of claims 1 - 10.

11. Method for detecting a louse on a fish, comprising the steps of:

- receiving an image of a fish;

- increasing a lice counter for each louse found in the image.

12. Method according to claim 1 1 , where the method is comprising the step of:

- computing the average number of lice per fish.