JP2013253868A - Method for measuring size of arbitrary portion of animal photographed by camera trap - Google Patents

Abstract

PROBLEM TO BE SOLVED: To measure a size of an animal from an image taken by a single camera.SOLUTION: A measuring method includes: a step for expressing a photographing space of a place for setting a camera trap as three-dimensional coordinates from a plurality of images obtained by photographing the place for setting the camera trap from a plurality of angles by using a three-dimensional model analysis method; a camera trap step for automatically photographing an animal entering into the photographing space by a single camera; a step for specifying a position in the photographing space by collating the same point between the image used for the three-dimensional model analysis and a camera trap image photographed by the camera trap step and performing three-dimensional processing of the camera trap image by applying the three-dimensional coordinates of the photographing space to the camera trap image; and a step for measuring a size of an arbitrary portion of the animal by forming the coordinates of a spatial position of the animal from a point on which the animal in the camera trap image touches the ground.

Description

  The present invention relates to an animal data acquisition method using a camera trap method. More specifically, the present invention relates to a method for measuring the size of an arbitrary part of an animal photographed by a camera trap using a single camera. In this specification, the camera trap or the camera trap method means that an infrared sensor is attached to the inside or outside of a single camera, and a wild animal passing in front of the camera is detected and automatically recorded. It is used to mean the technique to do.

  In wildlife surveys conducted in ecosystem conservation and management and environmental impact assessments, the use of the camera trap method is becoming increasingly popular as a method for effectively acquiring wildlife data that is rarely seen by humans (non- Patent Document 1). This camera trap method is based on a small single camera and can be easily installed, so it is currently being used in practical situations such as natural environment conservation and management and environmental assessment. It is out. For example, it is possible to detect changes in species, individual density, etc. by continuously installing cameras using the camera trap method and continuing shooting for a certain period of time, so various problems for recent wild animals such as agriculture and forestry by birds and beasts The camera trap method is positioned as an efficient monitoring method in terms of damage, ecosystem disturbance due to introduced species, biodiversity and the effects of global warming on the ecosystem.

  There are patterns, forms, and body sizes in the information of wild animals captured in images by the camera trap method. Of these, the body size information is the most basic quantitative data of individuals and species, but the size and size of the animal cannot be determined from a single camera image (2D) alone. Not available for For this reason, in the conventional camera trap method, only information on the type and frequency of imaging of the captured animal is used, and quantitative information on the body part in the image is useful for evaluation and management of animal populations. It is not utilized even though it is a lot of information. Information on the body size of an animal photographed with this camera trap cannot be obtained without capturing the animal.

  On the other hand, if body size information, which is the most basic quantitative data of individuals and species, can be obtained, it is possible to distinguish species and individuals that are difficult to distinguish from their forms and patterns, and the state and changes of animals and animal populations. The need for size measurement is high because it is possible to evaluate and manage the size. Therefore, conventionally, attempts have been made to make it possible to measure the body size of an animal only from a single camera image (two-dimensional). For example, in the case of bat surveys in Japan, by capturing the shadow of a bat on the ground, the approximate size of the part of the bat is measured, and species that are difficult to distinguish from the form alone are determined ( Non-patent document 2). Also, in the example of whale sharks in Western Australia, a major diver who swims in parallel with the sharks was photographed together with the sharks, and the size of the sharks was estimated from the comparison with the majors. The characteristics of the population, such as the regression relationship with dorsal fin), are grasped (Non-patent Document 3). Furthermore, a so-called 3D camera imaging technique (Non-Patent Document 4) in which a plurality of cameras capture an object simultaneously has been proposed.

Bookhout, T. A. 1996. Research and management techniques for wildlife and habitats.Allen Press, Inc. Hirakawa, H. and K. Maeda. 2006. A technique to estimate the approximate size of photographed bats. Wildlife Society Bulletin 34: 413-418. Meekan, MG, CJA Bradshaw, M. Press, C. McLean, A. Richards, S. Quasnichka, and JG Taylor. 2006. Population size and structure of whale sharks Rhincodon typus at Ningaloo Reef, Western Australia. Marine Ecology Progress Series 319 : 275-285. Harvey, E., M. Cappo, M. Shortis, S. Robson, J. Buchanan, and P. Speare. 2003. The accuracy and precision of underwater measurements of length and maximum body depth of southern bluefin tuna (Thunnus maccoyii) with a stereo-video camera system. Fisheries Research 63: 315-326.

  However, the invention described in Non-Patent Document 3 is effective for animals such as whale sharks that are not afraid of humans, and it is virtually impossible to shoot with a major together with wild animals that avoid humans. . Further, as in the invention described in Non-Patent Document 2, even if an attempt is made to capture a shadow of a wild animal with a camera trap, the spatial position and orientation of the animal cannot be accurately specified, and therefore size measurement cannot be performed with high accuracy. In addition, the so-called 3D camera photographing technique in which a plurality of cameras described in Non-Patent Document 4 simultaneously photograph an object cannot be realized by a small single camera, and the cost is high. Losing the convenience of the largest element that is. That is, it is not suitable for use in a scene where a conventional camera trap is used, and the usable scene is greatly restricted.

  It is an object of the present invention to provide a camera trap method that enables animal size measurement from a single camera image.

  In order to achieve this object, a method for measuring the size of an arbitrary part of an animal photographed with a camera trap according to claim 1 is a three-dimensional model analysis based on a plurality of images obtained by photographing a place where the camera trap is set from a plurality of angles. A method of converting a shooting space where a camera trap is set using a technique into a three-dimensional coordinate, a camera trap step of automatically shooting an animal that has entered the shooting space with a single camera, and the three-dimensional model analysis. The same point of the used image and the camera trap image taken in the camera trap step are collated, and the three-dimensional processing of the camera trap image is performed by applying the three-dimensional coordinates of the shooting space to the camera trap image. Giving the position in the shooting space, and a point where the animal in the camera trap image touches the ground So that and a step of measuring the size of any portion of the animal by coordinate the spatial position of the Luo animals.

  The invention described in claim 2 is the method for measuring the size of an arbitrary part of an animal photographed by the camera trap according to claim 1, wherein a measurement test measure is set at the photographing location. And

  The invention described in claim 3 is the method for measuring the size of an arbitrary part of an animal photographed by the camera trap according to claim 1 or 2, wherein a mark for point matching is set at the photographing location. It is characterized by.

  According to the first aspect of the present invention, the imaging space is three-dimensionally coordinated by three-dimensional model analysis using images obtained by photographing the same place as the camera trap method from a plurality of angles, and used for this three-dimensional model analysis. The spatial position of the animal from the point where the animal is in contact with the ground by collating the same point of the captured image and the same point of the image by the camera trap method and performing the three-dimensional coordinate processing of the two-dimensional image obtained by the camera trap method Can be coordinated, so that it is possible to measure the size of an arbitrary part of an animal photographed by a camera trap. Therefore, from the image of the animal photographed by the camera trap method, not only the pattern and form of the animal but also the information of the body size which is the most basic quantitative data of the individual and the species can be obtained.

  The inventors set a measurement test measure with a different length at the shooting location, and measured the length from the image captured by the camera trap method. As a result, the error was an average of 2%, and the measurement was performed with a certain degree of accuracy. I found that it was possible. Also, in the measurement of antelopes and sika deer taken by the camera trap method, the body size distribution is estimated from the identification of the position in three dimensions, the measurement of the height of hind legs and torso (Fig. 2), and the length of hind legs of multiple individuals. The usefulness was confirmed (FIG. 3). Therefore, the present invention can be applied to the analysis of detailed individuals and populations using size measurement in addition to the appearance frequency in the conservation and management of animals, particularly wild animals.

  According to the second aspect of the present invention, since a photographed image includes a measurement test measure whose dimensions are known, three-dimensional model analysis is facilitated.

  According to the invention described in claim 3, since the mark for point matching is included in the photographed image, it is set, so the matching of the same point between the image used for the three-dimensional model analysis and the camera trap image is performed. Becomes easy.

It is a flowchart which shows an example of embodiment of the measuring method of the size of the arbitrary site | parts of the animal image | photographed with the camera trap of this invention. It is a flowchart which shows an example of an image three-dimensionalization process. It is explanatory drawing which shows the example of body size measurement of a sika deer, an upper left figure shows the position in the imaging | photography space of the sika deer specified by coordinate conversion, and a lower right figure shows a camera trap image. It is explanatory drawing which shows the example of a measurement of the body size of an antelope, and shows the measurement example of the antelope ear and hind leg length after three-dimensional coordinate conversion. It is a graph which shows the example of the hind leg length size measurement of a sika deer, the vertical axis shows the range of the measured hind leg length (maximum value-minimum value), and the black circle shows the average value.

  Hereinafter, the configuration of the present invention will be described in detail based on embodiments shown in the drawings.

  FIG. 1 shows an example of an embodiment of a method for measuring the size of an arbitrary part of an animal photographed with a camera trap of the present invention (hereinafter abbreviated as a size measuring method using a camera trap). This size measurement method using a camera trap is a method of measuring the body size of a target animal from a two-dimensional image taken by a single camera. However, the animal to be photographed exists in a three-dimensional space, and when the image is photographed at a distance close to the camera or far from the camera, the image is photographed sideways with respect to the camera, the image is photographed obliquely, etc. In the above measurement, the size changes greatly. Therefore, if there is no information on the position of the target animal in the three-dimensional space (position having x, y, z coordinates), it is impossible to measure the size of the animal with a certain accuracy.

  Therefore, in the size measurement method using the camera trap of the present invention, the shooting space of the place where the camera trap is set is converted into a three-dimensional coordinate using a three-dimensional model analysis method from a plurality of images obtained by shooting the place where the camera trap is set from a plurality of angles. A two-dimensional camera trap image obtained by automatically photographing an animal that has passed in front of a camera trap camera (referred to as a camera trap image in this specification) and an image used for three-dimensional model analysis The camera trap image is subjected to three-dimensional processing by applying the three-dimensional coordinates of the shooting space to the camera trap image, and the position in the shooting space is specified. From the point where the animal in the camera trap image is in contact with the ground, The size of an arbitrary part of an animal can be measured by converting the spatial position into a coordinate.

  Specifically, as shown in FIG. 1, the same place as the camera trap method is photographed from a plurality of angles in advance or after the target in the same space as the camera trap method (step 2 a). Here, photographing is performed from a plurality of angles, for example, 3 to 5 directions so that the angles are different from each other. At this time, since it is preferable to improve the measurement accuracy that the same coordinates are taken with three or more photographs, for example, three cameras are arranged at an angle of 45 ° and photographed. In addition to taking pictures in the horizontal direction as well as changing the height, the measurement results can be improved. Therefore, it is preferable to change the height and further shoot with one or two cameras. Then, the imaging space of the place where the camera trap is set is converted into a three-dimensional coordinate from a plurality of images obtained by shooting the place where the camera trap is set from a plurality of angles (step 2b). At the time of the three-dimensional coordinate conversion, all other parts are set to the actual size by specifying the element / line having a known size in the captured image and inputting the actual size. This three-dimensional coordinate processing of the photographing space may be performed once for one photographing space, and can deal with a large number of animal images obtained by the camera trap method in the same photographing space. In addition, the three-dimensional coordinate processing of the photographing space may be performed before or after the camera trap method. Therefore, if the shooting space of the place where the camera trap is set is converted into three-dimensional coordinates, the past image assets shot in the shooting space and accumulated so far can be analyzed.

  On the other hand, a camera for camera trap is installed at the photographing location, and the animal passing in front of the camera is automatically photographed in two dimensions (step 1a), and an image showing the animal is accumulated (step 1b). Here, the shooting angle of the camera trap camera may be taken from the same angle as one of a plurality of images used for the three-dimensional model analysis, or from a different angle from any of the images. Also good. The camera trap method and the three-dimensional coordinated camera are preferably the same model, but need not be the same model, and may be a digital camera or a film camera.

  Further, the same point of the image by the camera trap method and the image used for the three-dimensional model analysis are collated, and the image of the camera trap method is converted into a three-dimensional coordinate (step 2c). Thereby, since the spatial position of the animal can be coordinated from the point where the animal is in contact with the ground, the size of an arbitrary part can be measured (step 2d).

(Process of 3D coordinateization of camera trap method image)
(photograph)
FieldNote DS6010 (manufactured by Marifu Shoji Co., Ltd., width 13.5 cm × height 12 cm × depth 7 cm) was used as a photographing apparatus by the camera trap method (FIG. 1, step 1a). Although not shown, this device is composed of a waterproof digital camera, μTOUGH-6010 manufactured by Pentax, and a camera case incorporating an infrared sensor. When the infrared sensor mounted on the top of the case reacts to wild animals, the switch on the case physically presses the power button of the digital camera, and the shutter is automatically pressed. Although this model was used in this report, as described in (2.1) above, the camera can be used with any digital camera or film camera regardless of the model, and an external infrared sensor is used for the camera on hand. It can also be applied to original equipment.

  The test site for photography was set in a coniferous forest near the altitude of 1800m in the lower part of Mt. The Mt. Fuji sub-alpine zone is a precious habitat for various wild animals (Ayumi Kobayashi et al. 2009) and is suitable for shooting various animal species. In addition, the subalpine zone is the highest altitude of the main wildlife habitat, but severe environmental conditions exist. If the usefulness and effectiveness of this method are verified at this point, we can make a certain evaluation regarding the practicality in many other regions in Japan.

  In selecting the camera location at the test site, the target point is selected in advance from the altitude and inclination on the map, and then the area around the target point and a radius of about 500 m are surveyed on the site. Or, I chose a space that doesn't block the front near the kemono road. Using a trunk such as dead trees, FieldNote DS6010 was set with a simple band at a height of about 1.5m from the ground.

  The above-mentioned μTOUGH-6010 is used as the same camera as the digital camera used in the camera trap method (FIG. 1, step 1a) for photographing (FIG. 1, step 2a) for converting the image of the camera trap method into three-dimensional coordinates. It was. The images were taken by the camera trap method, or toward the space where the images were photographed, so that overlapping locations could be seen from multiple angles with respect to the space.

(analysis)
FIG. 2 shows analysis items based on the concept of three-dimensional coordinate conversion and size measurement of captured images. The analysis items here, that is, the work from 3D model analysis to body size measurement,
PC software Photomodeler 2012 (Eos Systems Inc. 2012, http://www.photomodeler.com/contact_us/default.htm) was used. This software has a function of creating a three-dimensional model from a plurality of images and efficiently performing surveying from the images. It also has a function of easily performing camera lens calibration (calculation of camera lens distortion characteristics) necessary for pre-processing. Note that the measurement method according to the present invention can be implemented by other software that performs a three-dimensional model analysis from a plurality of photographs or by a self-made program if processing based on the basic concept described above is performed.

(3D modeling analysis)
A plurality of images were taken from different shooting angles at the shooting location of the camera trap method (FIG. 2, step 2b-1), and set as the same point on the image (a point that can be easily discriminated on the image, as a mark) (Markers etc.) are collated between images (about 10 to 15 points in the example of measurement at the test site by the PC software described above), and length information already measured in the imaging space is used (FIG. 2, step 2b-2) Three-dimensional model analysis based on triangulation is performed to determine the three-dimensional coordinates of the imaging space that is the target of the camera trap method (FIG. 2, step 2b-3). In the work in this report, in order to improve the efficiency of collating the common parts in the image, a small marker serving as a marker is placed on the ground beforehand so that it can be collated easily.

(3D coordinate)
An image of a wild animal that has entered the shooting space is captured by the camera trap method (FIG. 2, step 2c-1). Then, the camera trap image is collated with the three-dimensional space image obtained in step 2b-3 described above, and the camera trap image is converted into a three-dimensional coordinate (FIG. 2, step 2c-2). Since the aerial image has already been three-dimensionalized in step 2b-3, the collation work can be performed with as few as four or five points. If the settings of the photographed camera (that is, the angle of view) are the same during the camera trap method, the image collation is only required once, and the same coordinate information can be applied to other animal photographs. The xy coordinates of a single image taken of a wild animal are matched with the three-dimensional x, y, z coordinates (FIG. 2, step 2c-3).

(Space location and size measurement)
FIG. 3 shows a three-dimensional coordinated wild animal image. In this image, by creating a plane perpendicular to the ground on the line connecting the front foot and the hind paw where the animal has feet on the ground, the surface is parallel to the trunk. Measurement was possible (step 2d-1). Candidates for size measurement as a part in a plane parallel to the torso are the torso, the torso length, the shoulder height, the torso height, the forefoot, the back leg length, and the like. Note that the size measurement operation can be performed by designating a measurement target on the screen (step 2d-2).

(Image measurement accuracy verification and size measurement example for wild animals)
For the purpose of verifying the measurement accuracy in the image, a measure having a certain length (5 cm to 50 cm up to a distance of 5 cm, each two in length) is randomly placed in the shooting space. The measurement and actual length were compared. Furthermore, in order to examine the usefulness of the application to wild animals, we actually photographed the wild animals with the camera trap method, converted the obtained animal images into three-dimensional coordinates, and measured the body size of the animals. It was.

(result)
(Image measurement accuracy)
The average of the measured value error (compared with the actual measured value) obtained from the image is 0.51 cm (0.14-0.96 cm), and the average error accuracy is about 2%. It turns out that the object of the place can be measured.

(Size measurement for wild animals)
The example applied to the photographed sika deer is shown in FIG. A sika deer photographed on a two-dimensional plane was given a position and direction in a three-dimensional space by performing the three-dimensional coordinate conversion of the measurement method according to the present invention. As a result of image measurement, it was possible to measure parts such as hind legs and torso height in a plane parallel to the torso of the body. The same measurement was performed for the Japanese serow, and the result of measuring the back leg and the ear region is shown in FIG.

  About 80% of the animal images actually photographed were sika deer, so we attempted to measure multiple sika deer. The rear leg length (from the knee of the hind leg to the heel, see FIG. 3 and FIG. 4) was unified as the measurement site, which makes it easy to maintain a straight line of posture. From the image, we measured the male and female adults and those that can be judged as cubs by male horns etc. (Fig. 5). Each group's numerical value has little overlap with other groups. It was a thing. The accuracy cannot be verified without actually capturing these sika deer, but the measured values of the relevant parts of the sika deer actually captured in the past literature (including the hoof part: Takatsuki 1997, Naruki Takatsuki 2006) are the length of the hoof. (Kadozaki 2009) is a reasonable figure when compared.

(Applicability of this method)
In the present invention, in a camera trap method using a single camera, which has been used in the assessment research and research fields, the two-dimensional image is converted into a three-dimensional coordinate, thereby identifying the position of the photographed animal individual. Size measurement is possible. In the camera trap method that has been increasingly used as a simple method, in addition to information on the appearance frequency of animals in an image, information on the spatial position and the length of a body part can be acquired. There has been no way to obtain body size data other than relying on measurements from captured or dead individuals. Capturing animals is labor intensive, has a significant impact on individuals, and can be injured or killed by traps or anesthesia guns (Bookhout 1996). For this reason, methods that have little impact on animals are particularly needed in surveys related to business activities that require sociality. The method for measuring the size of any part of an animal photographed with a camera trap according to the present invention can meet practical needs in environmental impact assessment and wildlife management / conservation accompanying business activities, etc. The camera trap method equipment, which is already in use, is used and is considered to be a highly versatile technology.

  The size of the body obtained by the measurement method according to the present invention is one of the most basic information for evaluating an individual, an individual group, and a species. In addition, if body size measurement can be performed from an image by the measurement method according to the present invention, the size information determines the species classification in the case where the shape is similar between species or when it is difficult to distinguish the species from the image. It can be applied as a method. Furthermore, in the measurement example of sika deer in the measurement method according to the present invention, size information by male and female pups and adult animals was obtained, but by increasing these measurement data, qualitative characteristics within the species of animals, that is, It is also possible to analyze the properties (size distribution) as a population. Does the population tend to increase if there is information on the size class (child, young, adult, etc.) based on the size measurement by the measurement method according to the present invention in addition to the appearance frequency information by the original camera trap method? It is also possible to analyze population dynamics, such as whether there is a risk of extinction by declining. For example, whether there are increasing numbers of alien species that invade certain areas such as private land and national parks, whether conservation measures for important species implemented in land development have been properly taken, and populations have been restored. There is a possibility that it can be grasped only by the camera trap method.

  In addition, because of the widespread use of digital cameras, all data handled at present is digitized, so by introducing technologies such as automatic image transfer and image recognition by wireless communication, data on a point-by-point basis, such as maps, etc. It is possible to develop into evaluation. In practical application situations, for example, when quantitatively assessing rare species and important species inhabiting the region and assessing impacts, what kind of land these species are common in, and how much influence is due to the development of the planned land In many cases, a face-to-face evaluation is required. If the analysis results at each point can be mapped as plane information using the new camera trap method that performs three-dimensional coordinates, specific conservation, management, and development based on the characteristics of animal populations and trends Applicable to formulation.

  The above-described embodiment is an example of a preferred embodiment of the present invention, but is not limited thereto, and various modifications can be made without departing from the scope of the present invention. For example, in the present embodiment, the case where the present invention is applied to the camera trap method mainly used for wildlife surveys has been described. However, the present invention is not limited to wildlife surveys, and is not limited to wild animals such as domestic animals, other animals, and humans. It can also be used to understand and measure animal behavior. In addition, when a camera trap image with good shooting conditions can be obtained, one photo of the shooting space where the camera trap was placed before or after the camera trap was taken from an angle different from the camera trap camera position. Alternatively, if two or more sheets are prepared, a three-dimensional model analysis can be performed using the photograph and the camera trap image to perform a three-dimensional process of the camera trap image. Also in this case, the size of an arbitrary part of the photographed animal can be measured.

Claims (3)

3D coordinates the shooting space of the place where the camera trap is set using a 3D model analysis technique from a plurality of images obtained by shooting the place where the camera trap is set from a plurality of angles;
A camera trap step of automatically photographing an animal that has entered the photographing space with a single camera;
The camera trap image is obtained by collating the same point between the image used for the three-dimensional model analysis and the camera trap image photographed in the camera trap step, and applying the three-dimensional coordinates of the photographing space to the camera trap image. Performing a three-dimensional process and specifying a position in the imaging space;
Measuring the size of an arbitrary part of the animal by coordinating the spatial position of the animal from a point where the animal in the camera trap image contacts the ground. To measure the size of any part of an animal 2. The method for measuring the size of an arbitrary part of an animal photographed with a camera trap according to claim 1, wherein a measurement test measure is set at the photographing location. 3. The method for measuring the size of an arbitrary part of an animal photographed with a camera trap according to claim 1, wherein a mark for point comparison is set at the photographing location.

Images