Method and system for counting avian parasites - Patents.com

Abstract

1. A method for counting avian parasites (22) by capturing an image of an area of ​​interest (20) through which the parasites are expected to traverse and identifying said parasites using image recognition techniques, said area of ​​interest (20) being characterised as being part of a substrate (12) in which birds (10) are kept and having a topography with low temporal variation, said method comprising the step of counting occurrences of temporary, localised disturbances of the topography of the area of ​​interest.

Description

[Technical field]

[0001] The present invention relates to a method and system for counting bird parasites by capturing images of an area of ​​interest where the parasites are expected to traverse and identifying the parasites using image recognition techniques.

[0002] More particularly, the present invention relates to a method for detecting chicken mite infestations in poultry farms. [Background technology]

[0003] During the day, mites tend to hide in dark places such as cracks and crevices in the barns where poultry are kept. At night, when it gets dark and the chickens are resting on their perches, the mites crawl up to the chickens and suck their blood. Depending on the level of infestation, the blood loss caused to the chickens can be significant and detrimental to the chickens' health, resulting in a reduced growth rate or a reduced quality of the chickens' eggs. In either case, the mites cause significant losses to the poultry industry.

[0004] Established methods of pest control include mixing certain chemicals that kill mites into the drinking water for chickens. However, these measures are typically only taken when the barn is found to be infested. The present invention therefore aims to detect infestations as early as possible.

[0005] Detecting parasites by electronic image recognition is well known in the art. For example, machine learning techniques can be used to distinguish mites from the background, which may be, for example, the skin of an infested animal. However, in the case of bird or chicken infestations, it is more convenient to detect the mites as they crawl over the substrate on which the birds are kept. The problem with this approach is that the surface of the substrate, for example a wooden perch on which the chickens sit, has a relatively rough texture, which makes it difficult to distinguish the parasite from the background, especially when the image is taken at low light levels so as not to disturb the sleeping birds. It is therefore common practice that the area of ​​interest in which the image is captured is the floor of a box- or funnel-shaped detection device, which is placed in the path of the parasite and which constitutes a known, preferably uniform, background that contrasts well with the parasite. An example of this type of device is described in EP 2 931 032.

[0006] However, installing such detection devices in suitable locations is relatively expensive, and in particular care should be taken to avoid the formation of crevices between the detection device and the substrate on which it is installed, as mites would otherwise tend to crawl along those crevices and thereby avoid the floor of the detection device. Summary of the Invention [Problem to be solved by the invention]

[0007] It is therefore an object of the present invention to provide a low-cost yet efficient method for counting avian parasites. [Means for solving the problem]

[0008] To this end, the method according to the invention is characterized in that the area of ​​interest is part of a substrate on which birds are kept and which has a topography with low time variation, the method comprising a step of counting the occurrences of temporary local disturbances of the topography of the area of ​​interest.

[0009] The invention makes use of the fact that the parasites are crawling, i.e. moving, over the area of ​​interest, and therefore the disturbances caused by the crawling parasites at a given position of the substrate are only temporary. Despite the low contrast between the parasites and the background, these temporary disturbances can be easily detected by comparing images taken at different times. However, this method requires that the substrate itself has a stable topography over time, i.e. does not undergo substantial changes from one image to the other, typically stable for a period of up to 12-24 hours (this corresponds to the term "low time variation"). This requirement may not be met by a substrate consisting of, for example, mulch (which may be stirred by chickens). However, it would be met by a substrate constituted, for example, by a wooden perch, where the changes in the topography are no more than a gradual accumulation of dirt and dust on the surface and the occasional appearance of new scratches caused by the claws of the chickens.

[0010] In another aspect, the object of the present invention is achieved by a system configured to carry out the above-mentioned method.

[0011] More specific optional features of the invention are set out in the dependent claims.

[0012] In one embodiment, images of the area of ​​interest may be taken in the form of short video sequences that allow direct detection of the crawling parasite's movements. In another embodiment, the images may be composed of individual frames taken at longer time intervals. In that case, the crawling mites cause a local disturbance at a specific location in one image, while this disturbance is not visible in the next image as the mites move.

[0013] Depending on the average crawling speed of the mites and the speed at which the images are taken, it may happen that the mites are detected in multiple subsequent images, so that the count must be corrected for such double or multiple counts to obtain a valid measure of the degree of infestation. Nevertheless, it may be advantageous to use a very high image capture rate, where the distance traveled by the mites from one image to the other is relatively small, but significantly larger than the dimensions of an individual mite. This allows the movement of the mites to be safely tracked and a valid count to be obtained. This method has the additional advantage of increasing sensitivity due to the redundancy of repeated detection.

[0014] In order not to disturb the chickens, a relatively low level of illumination can be used in conjunction with a long exposure time for capturing the images. Local disturbances are then somewhat blurred by tick movements. However, this can even be an advantage, as disturbances are more easily visible as locations of reduced contrast in the contrast-enhanced image.

[0015] To improve the discrimination between mites and the background, it may also be useful to generate a reference image by overlaying several images taken at different times. Due to the motion of the mites, the overlay process only enhances the background features and not the mites, so that the reference image finally consists of almost pure background. Then, when this background image is subtracted from the captured image, the background becomes almost invisible and the disturbance (the mites) appears very clearly.

[0016] The method according to the invention requires only the installation of a camera at a suitable position, thus significantly reducing installation costs. However, it is possible to combine the camera with other sensors in order to obtain a deeper insight into the amount, condition and mechanism of the infestation. Examples of additional sensors include temperature sensors (e.g. for determining the activation time of the counting device and / or for studying the effect of light intensity on the mite behavior), humidity sensors, air pressure sensors, light intensity sensors. Position and / or acceleration sensors can be provided to detect any possible changes in the positioning and orientation of the camera. Acoustic sensors can be provided to record noises made by, for example, chickens in order to detect whether this noise correlates with mite activity.

[0017] The method and system according to the invention can provide farmers with an early warning in case of infestation. Besides this, the method and system can be used to document the time evolution of infestation and provide a simple gauge to assess the extent of infestation. These data can then be further used to correlate the extent of infestation with environmental conditions and / or chicken growth rate or other indicators of chicken health.

[0018] If multiple systems according to the invention are installed in the same or different barns, possibly even in different farmers' barns, it is also possible to collect statistical data showing how and from where the infestation spreads and which factors enhance or suppress the infestation.

[0019] Example embodiments will now be described in conjunction with the drawings. [Brief description of the drawings]

[0020] [Figure 1] 1 is a schematic perspective view of a counting system according to the present invention; [Diagram 2] 2 is an example of an image captured by the counting system according to FIG. 1; [Diagram 3]2 is an example of an image captured by the counting system according to FIG. 1; [Figure 4] 4 is an example of a reference image obtained by superimposing a plurality of images of the type shown in FIGS. 2 and 3. FIG. [Diagram 5] This is an image obtained by subtracting the reference image of FIG. 4 after image capture. [Figure 6] 2 is a flow diagram of a method according to the present invention; [Figure 7] FIG. 1 is a block diagram of a system according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Figure 1 shows a chicken 10 sitting on a wooden perch 12 on which it sleeps at night. A parasite counting system 14, comprising at least a digital camera 16 and a processing device 18, is installed in a suitable position to monitor a specific target area 20 on the perch 12. The camera 16 has an integrated lighting system for illuminating the target area 20 with visible or infrared light, especially at night when mites 22 tend to crawl along the perch to attack the chickens. The processing device 18 is configured to analyze the images captured by the camera 16 and to identify and count the mites 22 present in the target area 20 at the time the images were captured.

[0022] Figures 2 and 3 are examples of images A and B taken by camera 16 at different times. The two images A and B show essentially the same background 24 consisting mainly of the texture of the wooden surface of a perch within the area of ​​interest 20. Image A further shows four mites 22A that had crossed the area of ​​interest at the time the image was taken.

[0023] Image B also shows four ticks 22B in a different position than ticks 22A. Ticks 22B may or may not be identical to the four ticks 22A shown in image A. This depends on the time difference between the moment images A and B were captured.

[0024] 4 shows a reference image R obtained by overlaying images A and B, followed by renormalization of the image intensities. As a result, the features of the background 24, which are essentially the same in both images, appear enhanced, while the mites 22A, 22B are fainter. This overlay process can obviously be extended to many more images, with the result that the mites 22A, 22B and other mites that were only included in one image become almost invisible.

[0025] 5 shows an example of another image C captured at a later time than images A and B, from which a reference image R has been subtracted. As a result, the background 24 has been almost completely removed in image C, leaving only the three mites 22C captured in image C, as well as faint "ghosts" (i.e. negative images) of mites 22A and 22B. It will be appreciated that these ghosts will be even fainter if more than two images are superimposed.

[0026] Conventional image processing and / or machine learning techniques can then be used to assess the intensity and size of objects or disturbances visible in image C. Ghost images of mites 22A and 22B can be removed by comparing the intensity of these images to a threshold. The same is true for other disturbances, such as dust particles that have settled in the area of ​​interest between the capture of image B and image C, so that only mites 22C remain in image C. The dimensions of these local disturbances can be compared to upper and lower thresholds, and disturbances are counted as mites only if their dimensions are within reasonable limits. Thus, large area disturbances, i.e. the shadow of a chicken 10 falling on the area of ​​interest, are also removed. Mites 22C that pass the threshold test are then counted as a measure of the extent of infestation.

[0027] There are several strategies that can be used to avoid double or multiple counting. One strategy is to make the image capture rate very small so that it is possible to exclude that two images captured one after the other show the same tick. However, this can reduce the overall sensitivity of the system.

[0028] According to another approach, the capture rate is adapted to the average crawling speed of the mites, so that each tick crossing the area of ​​interest 20 is photographed, for example, three, four or five times. Then, by comparing the last three to five images, the movement of individual ticks can be tracked and the number of ticks that have crossed the area of ​​interest can be determined with high accuracy. This approach has the additional advantage that more information about the tick behavior, for example the average crawling speed, can be obtained, which can then be used to further optimize the algorithm.

[0029] FIG. 6 is a flow diagram of an example of a counting algorithm according to the present invention.

[0030] In step S1 an image counter n is initialized with n = 0. Then, in step S2, an image of the region of interest 20 is captured and saved, and the current content of the image counter, n, is assigned to that image.

[0031] Then, in step S3, the stored image is normalized.

[0032] In step S4 it is checked whether the image counter n (which is incremented later in the process) has already reached a value greater than 0. If so (y), then in step S5 a sliding average of the captured images is calculated. If n=1, the calculation of the sliding average need only consist of the superposition of the first two images as in Figures 2-4. Then, in the next execution of step S4, another image (n=3) is added, etc. When the superposition reaches a certain height, for example 10 images, in one embodiment it is possible to subtract the first image (n=0) from the superposition and add a new image instead, so that the superposition always includes the last 10 images.

[0033] In another embodiment, the first execution of step S5 may involve weighting the first image (n=0) with a certain weighting factor, for example 0.9, then adding the new image (n=1) with a weighting factor of 1.0, and then renormalizing the images to obtain a reference image R. Then, in subsequent executions of step S5, the previous reference image R is weighted with a weighting factor of 0.9, and each new image is added with its full weight. Thus, the reference image (sliding average) is always dominated by the last few images captured, while the information from the first few images (n=0, 1, ...) fades exponentially.

[0034] In step S6 it is checked whether the image counter n has reached a certain value n_min, averaged over a sufficient number of images such that the reference image is essentially "ghost-free". If that condition is met, in step S7 the reference image is subtracted from the number n-n_min images. In the first execution of this step, n is equal to n_min and the reference image is subtracted from image n=0, i.e. the first captured image is evaluated (retrospectively).

[0035] Next, in step S8, the disturbances remaining in the difference image (Image 0 - Image R) are checked against various intensity and size thresholds as described above, and any remaining disturbances found in the difference image may be subjected to a tracking routine to avoid double counting, after which the number of ticks is stored for that image.

[0036] In step S4, if it is determined that the value of n is 0, steps S5 to S8 are skipped. Similarly, in step S6, if it is determined that the condition is not satisfied, steps S7 and S8 are skipped.

[0037] Then, in step S9, it is checked whether a certain delay time has elapsed. It will be understood that this delay time defines the image capture rate. Step S9 is repeated until the specified delay time has elapsed, after which the image counter n is incremented by 1 in step S10 and the routine loops back to step S2. In this manner, the number of ticks can be determined and stored for each captured image, and the progression of the number of ticks over time can be stored and displayed.

[0038] Figure 7 is a block diagram of the processing device 18 shown in Figure 1. The input 26 of the processing system includes a camera interface 28 that receives image data from the camera 16. The input further includes a temperature sensor 30 for sensing the temperature of the direct environment of the perch 12, a humidity sensor 32 for sensing the air humidity of that environment, an air pressure sensor 34, a brightness sensor 36 for measuring the brightness of the illuminated target surface 20 (the brightness sensor may be integrated in the camera 16), a position and acceleration sensor 38 for detecting the position and possible movements of the whole counting system, and an acoustic sensor 40 for capturing chicken noises. The counting system or at least the cameras 16 may be installed on a rig that can be adapted to place the cameras at different positions around the perch 12, so that by referring to the information from the position sensor 38 it is possible to find out whether the mites prefer to crawl on the top or bottom surface of the perch. This information can then be utilized in further installations to optimize the camera positions.

[0039] A processing unit 42 processes the image data provided by the camera 14 as well as the sensor data from any other sensors in the input section 26 and stores the results in a memory 44, in particular the tick count history.

[0040] Statistical evaluation tools for evaluating the contents of memory 44 in different ways may also be implemented in the processing unit 42, allowing the tick count and sensor data to be subjected to various statistical analyses.

[0041] Furthermore, the data stored in memory 44, including the results of the analysis, may be transmitted to a communication section 46 which communicates with a user interface (not shown), for example a smartphone app, so that a user can retrieve the count and analysis results from memory 44. Furthermore, the processing unit 42 may implement an alarm system which, in case of first detection of a tick or other relevant event, can alert the user by sending a push message to the user interface.

Claims

1. A method for counting avian parasites (22) by capturing an image of a target area (20) expected to be traversed by a parasite and identifying the parasite using image recognition technology, wherein the target area (20) is part of a substrate (12) on which a bird (10) is housed and has a topography with little temporal variation, the method comprising the step of counting incidents of temporary local disturbances in the topography of the target area.

2. The method according to claim 1, wherein the target area is part of the surface of a perch (12).

3. The method according to claim 1, comprising the step of generating a reference image (R) showing the background (24) in the form of the texture of the substrate while suppressing other image features, and the step of suppressing the background (24) by subtracting the reference image from the captured image (C).

4. The capture speed at which the images (A, B, C) are captured is adapted to the average crawling speed of the parasite (22) such that each parasite crossing the target area (20) is captured several times, and the counting step includes the step of tracking the movement of the local disturbances representing the parasite. The method according to claim 1.

5. A system (14) for counting avian parasites (22), comprising a camera (16) and a processing device (18) arranged and configured to execute the method according to claim 1.

6. A system comprising: - a temperature sensor (30), - a humidity sensor (32), - a pressure sensor (34), - a light sensor (36), - a position and / or acceleration sensor (38), - at least one of an acoustic sensor (40), The system according to claim 5, wherein the processing device (18) is configured to associate the number of the parasites with the data provided by the sensors.

7. A software product comprising program code which, when executed on a processing device (18) of the system according to claim 5 or 6, causes the processing device to execute the method according to any one of claims 1 to 4.

Images