Processes of forming a panoramic image and apparatus therefor

Abstract

A process 30 of forming a panoramic image of a livestock habitat from images of different regions, whereby a panorama forming phase comprises: receiving adjacent images 32 representing different but overlapping regions of the habitat; identifying at least one common feature 34; determining a first location in a first image of each common feature and a second location in a second image of each common feature; and forming a trial panoramic image 36 in which the overlap between the first and second images is determined based on the first and second locations. An error addressing phase comprises: projecting onto an object plane of the habitat of at least one habitat feature in each image 40; determining an error between a location of each projected habitat feature and a pre-measured location of the corresponding habitat feature in a map of the object plane 44; and applying transformations to the images 46; wherein the trial image formed in the panorama forming phase providing the smallest combination of errors is chosen as the panoramic image 56. Also disclosed is an apparatus for carrying out said process.

Description

The present invention relates to processes of forming a panoramic image from plural images. The present invention also relates to apparatus for forming a panoramic image from plural images. Background Art It is known to monitor livestock by way of cameras. In an approach to livestock monitoring, cameras are installed at spaced apart and elevated locations and are oriented such that each camera acquires images from above of a different part of a livestock habitat, such as a bam. It is further known to instal the cameras such that adjacent cameras acquire images of different but overlapping parts of the livestock habitat to thereby provide for confidence in uninterrupted coverage of the livestock habitat. A panoramic image of at least a part of the livestock habitat is formed from plural images which are each acquired by a respective one of the cameras. In view of adjacent cameras acquiring overlapping images, an extent of overlap of adjacent images needs to be determined. This is typically done by identifying at least one feature in each of two adjacent images and which is common to the adjacent images and using the location of the at least one feature in each of the adjacent images to determine the extent of overlap. The present inventors have found that although the above described process of forming a panoramic image can provide sufficient registration of adjacent images where they overlap, there is often improper registration of adjacent images in respect of their non-overlapping parts. The effect of improper registration of adjacent images may be more noticeable when a larger number of adjacent images, such as a row or column of at least half a dozen images, are formed into a panoramic image whereby registration errors accumulate. Such a panoramic image formed of improperly registered images may not represent the livestock habitat properly. For example, and where the livestock habitat is a barn, the panoramic image of the barn may be generally trapezoidal when the barn is rectangular. The present invention has been devised in light of the inventors’ appreciation of the above problem. It is therefore an object for the present invention to provide an improved process of forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat. The present invention also relates to apparatus for forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat. Statement of Invention According to a first aspect of the present invention there is provided a process of forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat, the process comprising a panorama forming phase and an error addressing phase, the panorama forming phase comprising: a) receiving first and second adjacent images which represent different but overlapping parts of the livestock habitat; b) identifying at least one feature common to the first and second images; c) determining a first location in the first image of each common feature and a second location in the second image of each common feature; and d) forming a trial panoramic image by electronic methods in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations, the error addressing phase carried out using electronic methods and comprising: e) projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image; f) determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane, the location of the corresponding habitat feature in the map having been determined by measurement in the livestock habitat; and g) applying a first transformation to the first image and a second transformation to the second image and repeating steps d) to f), wherein step g) is performed at least once with each of the first transformations differing each application and each of the second transformations differing each application, and the trial panoramic image formed under step d) which provides the smallest combination of the location errors obtained from step f) is selected as the panoramic image. The process of forming a panoramic image of a livestock habitat, such as a barn, from plural images of different parts of the livestock habitat comprises a panorama forming phase and an error addressing phase. The panorama forming phase comprises receiving first and second adjacent images which represent different but overlapping parts of the livestock habitat. The panorama forming phase then comprises identifying at least one feature common to the first and second images and determining a first location in the first image and a second location in the second image of each common feature. Each of the first and second locations may be determined in respect of the respective one of the first and second images per se. For example, the first location may be determined as pixel a, b in the first image and the second location may be determined as pixel c, d in the second image. Thereafter the panorama forming phase comprises forming a trial panoramic image by electronic methods in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations. The error addressing phase is carried out using electronic methods and comprises projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image. For example, where the livestock habitat is a barn the object plane may correspond to the floor of the barn. The error addressing phase also comprises determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane, the location of the corresponding habitat feature in the map having been determined by measurement in the livestock habitat. The error addressing phase further comprises applying a first transformation to the first image and a second transformation to the second image and repeating the preceding steps from the step of forming the trial panoramic image. The step of applying the first transformation to the first image and the second transformation to the second image and repeating the preceding steps is performed at least once with the first transformations differing each application and the second transformations differing each application. Furthermore, the trial panoramic image which provides the smallest combination, such as the sum, of the location errors from the iterations is selected as the panoramic image. The present process addresses the above described problem of the known process by way of the error determining phase. More specifically, the above described problem is addressed by determining the error between the location of each projected habitat feature and the location of the corresponding habitat feature in the map of the object plane and repeating at least once such error determinations after application of first and second transformations. At least one habitat feature in each of the first and second images comprised in the trial panoramic image is projected onto the object plane of the livestock habitat. The first and second images represent different but overlapping parts of the livestock habitat. The parts of the livestock habitat represented in the first and second images may differ from each other in respect of translation and more specifically in respect of two or perhaps three components of translation. The part of the livestock habitat represented in each of the first and second images may be stored as coordinates, such as x, y and z coordinates, from an origin common to the first and second images. Each of the first and second images of the panoramic image may therefore be stored with coordinates whereby the panoramic image may be formed. The stored coordinates may be determined in dependence on the trial panoramic image which provides the smallest combination of the location errors. Furthermore, each of the first and second images, which represent different parts of the livestock habitat, may be characterised by rotation and more specifically by two or perhaps three components of rotation. Each of the first and second images of the panoramic image may therefore be stored with at least one component of rotation. When a first camera, which subsequently acquires the first image, and a second camera, which subsequently acquires the second image, are installed, their installation may be in accordance with a specification which is expressed in respect of components of translation and perhaps also at least one component of rotation. However, the specification may be incorrect and / or the camera may not be installed in accordance with the specification. Each of the first and second transformations may therefore change at least one component of translation and / or component of rotation of the image to which the transformation is applied. Each of the first and second transformations may thus be operative to algorithmically adjust the position of the first and second cameras as installed in accordance with the specification. The trial panoramic image formed under step d) which provides the smallest combination of the location errors obtained from step f) is selected as the panoramic image. A location error is determined for each projected habitat feature and at least one habitat feature in each of the first and second images is projected onto the object plane. This means at least two location errors are determined. The at least two location errors may be combined by summing the at least two location errors. The panorama forming phase comprises: receiving first and second adjacent images which represent different but overlapping parts of the livestock habitat; identifying at least one feature common to the first and second images; and determining a first location in the first image of each common feature and a second location in the second image of each common feature. In a first approach, the three steps of the panorama forming phase in the immediately preceding paragraph may be performed by electronic methods. More specifically, the step of identifying at least one feature common to the received first and second images may comprise detecting with an object detection algorithm an object common to the first and second images in each of the first and second images. To achieve sufficient accuracy in determination of the first and second locations, the common object should lie substantially in the same plane. Many objects, such as parts of the livestock habitat and apparatus in the livestock habitat, do not lie in the same plane. Therefore, the common object may be a planar design and more specifically a planar optically readable design, such as a QR code. An optically readable design, such as a QR code, may provide sub-pixel accuracy in determination of the first and second locations. Although an optically readable design may be advantageous from this perspective, the floor or ground in a livestock habitat is often bumpy whereby it may be difficult to have the optically readable design lie flat. In a second approach, the step of receiving first and second adjacent images which represent different but overlapping parts of the livestock habitat may comprise displaying on a display screen or printing the first and second adjacent images. The remaining two of the three steps of the panorama forming phase may be performed manually at least in part. The step of identifying at least one feature common to the first and second images may be performed by a person visually identifying the at least one common feature in the first and second images displayed on the display screen or in printed form. When the first and second images are displayed on the display screen, the step of determining a first location in the first image of each common feature and a second location in the second image of each common feature may comprise a person electronically determining the first and second locations in dependence on manual operation of an input device, such as operation of a mouse to move a cursor over the common feature and clicking of the mouse button to determine the location. When the first and second images are displayed in printed form, the step of determining a first location in the first image of each common feature and a second location in the second image of each common feature may comprise a person determining the first and second locations by measurement of the printed form of the first and second images. Irrespective of whether the first or second approach is followed, step d) comprises forming a trial panoramic image by electronic methods in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations. The process may further comprise applying a perspective transformation by electronic methods to each of the first and second images comprised in the trial panoramic image. In addition, the process may comprise visually inspecting the trial panoramic image after perspective transformation to determine how well the first and second images are registered. If the first and second images are clearly out of registration, steps b) to d) are repeated. Although use of the perspective transformation is optional, it may serve to provide for approximate registration of the first and second images to thereby reduce the computational burden imposed by the subsequent error addressing phase. More specifically, fewer different first and second transformations may need to be applied. According to the process, at least one habitat feature in each of the first and second images comprised in the trial panoramic image is projected onto an object plane of the livestock habitat. The object plane may lie at any level within the livestock habitat, although a plane, such as the ground, which is convenient having regard to the configuration of the livestock habitat may be chosen. Each at least one habitat feature may be projected onto the object plane by application of trigonometry and its location within the object plane determined. Trigonometry may be applied by electronic methods. The locations of the projected habitat features may be stored in a data store. The locations of the projected habitat features may be comprised in a data structure stored in the data store. According to the process, a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane is determined. The location of the corresponding habitat feature in the map has been determined before the process has started and by measurement in the livestock habitat. Measurement in the livestock habitat may be done by a person using one or more measurement approaches of known form and function. The object plane may lie at any level within the livestock habitat, although a plane, such as the ground, which is convenient having regard to the configuration of the livestock habitat and measurement thereof may be chosen. The map of the object plane may be stored in a data store. The map of the object plane may be constituted as a data structure stored in the data store. The location of the at least one corresponding habitat feature may be comprised in the data structure. The panorama forming phase comprises: identifying at least one feature common to the first and second images; and determining a first location in the first image of each common feature and a second location in the second image of each common feature. More specifically, plural features common to the first and second images may be identified, and their locations determined. Furthermore, where there are three or more plural common features, the plural common features may be spaced apart in two orthogonal directions whereby the common features do not lie along a straight line. In certain circumstances, such as when a perspective transformation is applied, there may be at least four common features. Alternatively or in addition, the plural features may be spaced apart such that they are located towards a periphery of the first and second images. Having plural features spaced apart in this way may provide for improved accuracy of overlap determination. The error addressing phase comprises projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image. There may be plural habitat features in each image. A habitat feature may be a feature common to the first and second images, such as the common feature used to determine the first and second locations. Alternatively or in addition, a habitat feature may not be common to the first and second images whereby the habitat feature is present in only one of the first and second images. Having at least one habitat feature which is not common to the first and second images may, in some circumstances, provide for improved registration of the first and second images in the panoramic image. The steps of forming a trial panoramic image and forming the panoramic image may be performed to an extent sufficient to allow for rendering, such as on a display screen or in printed form, of the trial panoramic image and the panoramic image. The steps may therefore be performed such that the first and second images are stored, such as in at least one data structure, with overlap data which determines their extent of overlap when the trial panoramic image and the panoramic image are rendered. The panorama forming phase may comprise acquiring the first image with a first camera and acquiring the second image with a second camera. The first and second cameras may be installed in or above the livestock habitat whereby each camera acquires from above an image of a respective part of the livestock habitat. Alternatively or in addition, each of the first and second cameras may be a 2-dimensional camera. A two-dimensional camera may provide superior performance, such as in respect of resolution, to a three-dimensional camera. Each camera may be an Internet Protocol (IP) camera. The process of forming a panoramic image may involve use of three or more cameras. More specifically, plural cameras may be installed in the livestock habitat in an array. Adjacent ones of the plural cameras in a column-wise direction and a row-wise direction may be subject to the panorama forming phase and the error addressing phase described above. The livestock habitat may be an indoors or outdoors habitat for quadrupedal livestock or poultry. The livestock habitat may be configured accordingly in respect of shape and / or size. Alternatively or in addition, the livestock habitat may be configured accordingly in respect of apparatus within the livestock habitat. For example and where the livestock animals are cattle, there may be a milking parlour within the livestock habitat. Steps of the process are performed by electronic methods. At least one such step may be performed by a processor. The processor may comprise at least one digital processor. According to a first approach, a digital processor may be disposed at each of plural cameras whereby at least one step of the process is carried out locally to a camera. The first approach may be applied together with at least one of the second and third approaches described below. According to a second approach, a digital processor may be at a location in the livestock habitat spaced apart from the plural cameras, such as on the ground inside a barn, the processor receiving images from the plural cameras. According to a third approach, a digital processor may be located remotely from the livestock habitat, the digital processor receiving images from the plural cameras. Furthermore, images may be conveyed from the plural cameras to the remotely located digital processor by way of the Internet. The digital processor of the third approach may be constituted by a cloud computing arrangement. The digital processor may therefore be of distributed form. The second and third approaches may be applied together with, for example, preliminary processing according to the second approach and further processing according to the third approach. The processor may be configured to perform one or more of the processes described herein. Apparatus may comprise structures and / or non-transitory memory having programmed instructions which are operated on by the processor and may further comprise electronic circuitry to perform these processes. Data of the process, such as the first and second images, the panoramic image, the map, and data derived therefrom, may be stored as electronic data in at least one data store. According to a second aspect of the present invention there is provided an apparatus for forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat, the apparatus comprising first and second cameras and a processor which is configured to control at least a part of a panorama forming phase and to control an error addressing phase, wherein the panorama forming phase comprises: a) receiving a first image from the first camera and a second image from the second camera, the first and second images adjacent and representing different but overlapping parts of the livestock habitat; b) identifying at least one feature common to the first and second images; c) determining a first location in the first image of each common feature and a second location in the second image of each common feature; and d) forming a trial panoramic image under control of the processor in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations, and wherein the error addressing phase comprises: e) projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image; f) determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane, the location of the corresponding habitat feature in the map having been determined by measurement in the livestock habitat; and g) applying a first transformation to the first image and a second transformation to the second image and repeating steps d) to f), wherein step g) is performed at least once with each of the first transformations differing each application and each of the second transformations differing each application, and the trial panoramic image formed under step d) which provides the smallest combination of the location errors obtained from step f) is selected as the panoramic image. Embodiments of the second aspect of the present invention may comprise one or more features of the first aspect of the present invention. Brief Description of Drawings Further features and advantages of the present invention will become apparent from the following specific description, which is given by way of example only and with reference to the accompanying drawings, in which: Figure 1 is a block diagram representation of apparatus according to an embodiment of the invention; and Figure 2 is a flow chart of a process according to an embodiment of the invention. Description of Embodiments A block diagram representation of apparatus according to an embodiment of the invention is shown in Figure 1. The apparatus is installed in a bam 10 (which constitutes a livestock habitat) which contains a milking parlour 12. The milking parlour 12 comprises plural stalls 14. For simplicity, Figure 1 shows a small number of stalls 14. In a practical embodiment, the milking parlour is of known herringbone or circular form and has many stalls. The apparatus comprises an array of cameras 16. The cameras 16 are installed at spaced apart locations in the roof of the barn and oriented downwards such that each camera acquires images of a different part of the barn whereby cattle can be recognised and monitored. The cameras 16 are installed such that adjacent cameras acquire images of different but overlapping parts of the barn to thereby provide for confidence in uninterrupted coverage of the bam. The extent of overlap of adjacent images depends on the livestock animals being monitored and the purpose of livestock monitoring. In the present embodiment, individual cows are monitored and tracked. To provide for proper tracking of a cow from image to image, the overlap of adjacent images is preferably one cow wide. The cameras 16 are installed in accordance with an installation specification that determines extent of coverage by the cameras and the overlaps of adjacent images. The location and orientation of each camera is therefore determined by the installation specification. The location of each camera is specified by way of x, y and z coordinates from a common origin in the barn, such as a corner of the bam. The orientation of each camera is specified by way of pitch, yaw and roll. The apparatus yet further comprises a central digital processor 18 of conventional form and function which is installed in a corner of the bam and which is in communication with the cameras 16. In a first form of the embodiment of the apparatus, image and data processing steps according to the invention are performed by a digital processor local to each camera 16. The local digital processor is a Jetson Xavier NX from NVIDIA, Trinity House, Cambridge Business Park, Cowley Road, Cambridge, CB4 OWZ, United Kingdom. Later processing steps, such as updating of records for cows tracked by the cameras, are performed in the central digital processor 18. In a second form of the embodiment, the apparatus comprises a cloud computing resource (not shown) of conventional form and function. The second form is as per the first form but with the later processing steps performed in the cloud computing resource instead of in the central digital processor 18. In a third form of the embodiment, each local digital processor is operative merely as a relay station to convey image and other data, which has been generated locally, to the cloud computing resource by way of the Internet or to the central digital processor 18 wherein steps according to the invention are performed. In a fourth form of the embodiment, substantially all image and data processing steps according to the invention are performed in a distributed fashion by way of a Jetson Xavier NX digital processor (not shown) local to each camera 16 and later processing steps, such as updating of records for cows tracked by the cameras, are performed in these local digital processors. A flow chart of a process 30 according to an embodiment of the invention is shown in Figure 2. The process 30 makes use of the apparatus shown Figure 1. The process in respect of an image acquired by each of two adjacent cameras 16 will now be described. This process is the same for images acquired by further pairs of adjacent cameras 16. The process comprises a panorama forming phase and an error addressing phase. The panorama forming phase starts with acquisition of a first image with a first camera 16 of the pair of adjacent cameras and of a second image with a second camera 16 of the pair of adjacent cameras 32. The next step in the panorama forming phase is identification of four features common to the first and second images 34. In a first approach to identification of four common features, four unique QR codes have been laid on the ground of the barn at widely spaced apart locations corresponding to near the periphery of the first and second images and within the overlapping parts of the first and second images according to the installation specification. Each QR code constitutes a respective one of the four common features. The QR codes are read by the digital processor from the first and second images. The location of each read QR code in the first image is then determined (with each of these four locations constituting a first location) and the location of each read QR code in the second image is determined (with each of these four locations constituting a second location) by the digital processor. Each determined location is expressed in respect of a corner of the QR code being at the M, N pixel of the respective camera. In a second approach to identification of four common features, the first and second images are displayed on a display screen or printed in hardcopy. When the first and second images are displayed on the display screen, determining of first and second locations of common features in the first and second images, such as a part of the bam apparatus or a puddle, comprises a person operating a mouse to move a cursor over the common feature and clicking of the mouse button to determine the location. Each determined location is expressed in respect of a point of the common feature being at the M, N pixel of the respective camera. When the first and second images are displayed in printed form, determining the first and second locations of common features in the first and second images comprises a person determining the first and second locations by measurement of the printed form of the first and second images. Again, each determined location is expressed in respect of a point of the common feature being at the M, N pixel of the respective camera. Irrespective of whether the first or second approach is followed, the panorama forming phase further comprises the digital processor forming a trial panoramic image in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations 36. The panorama forming phase further comprises the digital processor applying a perspective transformation to each of the first and second images comprised in the trial panoramic image 38. In addition, the panorama forming phase comprises visually inspecting the trial panoramic image after perspective transformation to determine how well the first and second images are registered. If the first and second images are clearly out of registration, the preceding steps of the panorama forming phase after acquisition of the first and second images are repeated. The perspective transformation serves to provide for approximate registration of the first and second images to thereby reduce the computational burden imposed by the subsequent error addressing phase. The process then moves on to the error addressing phase. The error addressing phase starts with the digital processor being operative to project plural habitat features in each of the first and second images onto an object plane of the barn 40. The object plane may lie at any level within the bam, although a plane, such as the ground, which is convenient having regard to the configuration of the bam is chosen. Each habitat feature is projected onto the object plane by application of trigonometry. Furthermore, the location of each habitat feature within the object plane is determined by operation of the digital processor 42. Application of trigonometry is within the ordinary design skills of the person skilled in the art. The locations of the projected habitat features are stored in a data store of the apparatus. A habitat feature is one of the features common to the first and second images and for which one of the first and second locations is determined during the panorama forming phase. Alternatively, a habitat feature is not common to the first and second images whereby the habitat feature is present in only one of the first and second images. Such a habitat feature is the like of a part of the barn apparatus or a puddle. In some circumstances, having at least one habitat feature which is not common to the first and second images provides for improved registration of the first and second images in the panoramic image. The error addressing phase further comprises the digital processor determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane 44. Prior to the start of the present process, such as when the cameras were installed in the barn, the locations of the corresponding habitat features in the map were determined by measurement in the bam. Bam measurements are made by a person using one or more measurement approaches of known form and function. As described above, the object plane lies at any level within the bam, although a plane, such as the ground, which is convenient having regard to the configuration of the bam and measurement thereof is chosen. The location of each corresponding habitat feature is converted to an appropriate form, such as a pixel address relative to an origin for the barn, and before the converted location is included in the map. Converting the location of each corresponding habitat feature to such an appropriate form provides for ease of determining the error between the location of each projected habitat feature, which is in the same form, and the location of the corresponding habitat feature in the map. The map of the object plane with locations of corresponding habitat features is stored in the data store. Thereafter, the error addressing phase comprises applying a first transformation to the first image and a second transformation to the second image 46. Each of the first and second transformations transforms the image to which the transformation is applied by changing at least one of the x, y and z coordinates, and the pitch, yaw and roll according to the installation specification for the camera 16 that acquired the image. Then a fresh a trial panoramic image is formed 48, plural habitat features are projected onto the object plane of the barn 50, and location errors for the projected habitat features are determined 52. Thereafter, a first transformation is applied to the first image and a second transformation is applied to the second image 46 and the process repeats steps 48 to 52 until plural sets of location errors are obtained with one set of location errors obtained per iteration. The first transformation differs each application and the second transformation differs each application. The location errors in each set of location errors are combined by summing and the combinations are compared with one another to identify the smallest combination 54. The trial panoramic image which has yielded the smallest combination of location errors is then selected as the panoramic image 56 for use in subsequent processes, such as cow monitoring and tracking. The process described above with reference to Figure 2 is carried out in respect of all adjacent pairs of images. Applying a transformation to an image in respect of one adjacent image will have an effect on registration of that image with at least one further adjacent image. Therefore, the process further comprises an optimisation 5 phase in which a combination for each pair of adjacent images is selected which provides for optimal minimum combinations across other adjacent images.

Claims

1. A process of forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat, the process comprising a panorama forming phase and an error addressing phase, the panorama forming phase comprising:a) receiving first and second adjacent images which represent different but overlapping parts of the livestock habitat;b) identifying at least one feature common to the first and second images;c) determining a first location in the first image of each common feature and a second location in the second image of each common feature; andd) forming a trial panoramic image by electronic methods in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations, the error addressing phase carried out using electronic methods and comprising:e) projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image;f) determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane, the location of the corresponding habitat feature in the map having been determined by measurement in the livestock habitat; andg) applying a first transformation to the first image and a second transformation to the second image and repeating steps d) to f), whereinstep g) is performed at least once with each of the first transformations differing each application and each of the second transformations differing each application, and the trial panoramic image formed under step d) which provides the smallest combination of the location errors obtained from step f) is selected as the panoramic image.

2. The process according to claim 1, in which the first location is determined as pixel a, b in the first image and the second location is determined as pixel c, d in the second image.

3. The process according to claim 1 or 2, in which the part of the livestock habitat represented in each of the first and second images is stored with coordinates from an origin common to the first and second images whereby the selected panoramic image is formed, the stored coordinates determined in dependence on the trial panoramic image which provides the smallest combination of the location errors.

4. The process according to claim 3, in which each of the first and second images, which represent different parts of the livestock habitat, is characterised by rotation, and each of the first and second images of the panoramic image is stored with at least one component of rotation.

5. The process according to any one of the preceding claims, in which first and second cameras are installed at respective positions in the livestock habitat in accordance with an installation specification, the first and second cameras respectively acquiring the first and second images, and each of the first and second transformations changes at least one component of translation and / or at least one component of rotation of a respective one of the first and second images to which the transformation is applied, whereby each of the first and second transformations is operative to algorithmically adjust the positions of the first and second cameras as installed in accordance with the installation specification.

6. The process according to any one of the preceding claims, in which at least two location errors are determined and the at least two location errors are combined by summing the at least two location errors.

7. The process according to any one of the preceding claims, in which steps a) to c) are performed by electronic methods.

8. The process according to claim 7, in which the common feature is an object common to the first and second images and step b) comprises detecting thecommon object in each of the first and second images with an object detection algorithm.

9. The process according to claim 8, in which the common object is a planar optically readable design.

10. The process according to any one of claims 1 to 6, in which step a) comprises displaying the first and second adjacent images on a display screen, step b) is performed by a person visually identifying the at least one common feature in the first and second images displayed on the display screen, and step c) comprises a person electronically determining the first and second locations by manual operation of an electronic input device.

11. The process according to any one of the preceding claims further comprising after step d) applying a perspective transformation by electronic methods to each of the first and second images comprised in the trial panoramic image, and repeating steps b) to d) if the perspectively transformed first and second images are out of registration by more than predetermined extent.

12. The process according to any one of the preceding claims, in which each at least one habitat feature is projected onto the object plane by application of trigonometry by electronic methods.

13. The process according to any one of the preceding claims, in which the map of the object plane is constituted as a data structure, the data structure is stored in a data store, and the location of the at least one corresponding habitat feature is comprised in the data structure.

14. The process according to any one of the preceding claims, in which step b) comprises identifying at least three features common to the first and second images, step c) comprises determining respective first and second locations in the first and second images, and the at least three common features are spaced apart in twoorthogonal directions whereby the at least three common features do not lie along a straight line.

15. The process according to claim 14 when depending from claim 11, in which step b) comprises identifying at least four features common to the first and second images and step c) comprises determining respective first and second locations in the first and second images.

16. The process according to claim 14 or 15, in which the at least three features are spaced apart such that they are located towards a periphery of the first and second images.

17. The process according to any one of the preceding claims, in which at least one habitat feature is not common to the first and second images.

18. The process according to any one of the preceding claims, in which the panorama forming phase comprises acquiring the first image with a first camera and acquiring the second image with a second camera, the first and second cameras installed in or above the livestock habitat whereby each camera acquires from above an image of a respective part of the livestock habitat.

19. The process according to claim 18, in which each of the first and second cameras is a 2-dimensional camera.

20. The process according to claim 18 or 19, in which plural cameras are installed in the livestock habitat in an array, and adjacent ones of the plural cameras in a column-wise direction and a row-wise direction are subject to the panorama forming phase and the error addressing phase.

21. The process according to any one of the preceding claims, in which the livestock habitat is an indoors or outdoors habitat for quadrupedal livestock or poultry.

22. The process according to any one of the preceding claims, in which the first and second images are acquired by a respective camera and a digital processor is at each of the cameras whereby at least one of steps a) to g) is carried out locally to each camera by the respective digital processor.

23. The process according to any one of the preceding claims, in which the first and second images are acquired by a respective camera, a digital processor is at a location in the livestock habitat spaced apart from the cameras, and the remotely located digital processor receives the first and second images from the cameras.

24. The process according to any one of the preceding claims, in which the first and second images, the trial panoramic image and the map are stored as electronic data in at least one data store.

25. An apparatus for forming a panoramic image of a livestock habitat from plural images of different parts of the livestock habitat, the apparatus comprising first and second cameras and a processor which is configured to control at least a part of a panorama forming phase and to control an error addressing phase, wherein the panorama forming phase comprises:a) receiving a first image from the first camera and a second image from the second camera, the first and second images adjacent and representing different but overlapping parts of the livestock habitat;b) identifying at least one feature common to the first and second images;c) determining a first location in the first image of each common feature and a second location in the second image of each common feature; andd) forming a trial panoramic image under control of the processor in which the first and second images are disposed relative to each other with their extent of overlap determined in dependence on the first and second locations, and wherein the error addressing phase comprises:e) projecting onto an object plane of the livestock habitat at least one habitat feature in each of the first and second images comprised in the trial panoramic image;f) determining a location error between a location of each projected habitat feature and a location of the corresponding habitat feature in a map of the object plane, the location of the corresponding habitat feature in the map having been determined by measurement in the livestock habitat; and5 g) applying a first transformation to the first image and a secondtransformation to the second image and repeating steps d) to f), whereinstep g) is performed at least once with each of the first transformations differing each application and each of the second transformations differing each application, and the trial panoramic image formed under step d) which provides the io smallest combination of the location errors obtained from step f) is selected as the panoramic image.28 10 24

Images