Three-dimensional cattle imager

Abstract

An in vivo animal retention, data collection and analysis system that comprises an animal retention structure that is structured and operable to temporarily restrain a live animal to be scanned in specific location. The system additionally comprises an animal data acquisition system that is structured and operable to acquire phenotypic data of the animal to be scanned. The system further comprises a data processing system that is structured and operable to execute one or more animal efficiency algorithm that utilizes the acquired phenotypic data to calculate the efficiency of the animal to be scanned.

Description

Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 THREE-DIMENSIONAL CATTLE IMAGER CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No.63 / 518,135, filed on August 8, 2023. The disclosure of the above application is incorporated herein by reference in its entirety. GOVERNMENT RIGHTS

[0002] This invention was made with government support under 2021- 67021-33448 awarded by the National Institute of Food and Agriculture. The government has certain rights in the invention. FIELD

[0003] The present teachings relate to systems and methods for determining various parameters of livestock utilizing 3-dimensional imagery, more particularly for determining the basal metabolic rate of cattle utilizing 3-dimensional imagery. BACKGROUND

[0004] The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.

[0005] The measurement of body weight of livestock has been frequently measured and used in metabolic research and prediction, especially in domesticated livestock. Further, in the last several years the beef industry has seen the launch of automated weight collection devices and services. However, weight is a poor indicator of basal metabolic rate, and is much less accurate and less consistent across species than surface area. SUMMARY

[0006] In various embodiments, the present disclosure provides an in vivo animal retention, data collection and analysis system that comprises an animal retention structure that is structured and operable to temporarily restrain a live 25957884.v1 - 1 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 animal to be scanned in specific location. The system additionally comprises an animal data acquisition system that is structured and operable to acquire phenotypic data of the animal to be scanned. The system further comprises a data processing system that is structured and operable to execute one or more animal efficiency algorithm that utilizes the acquired phenotypic data to calculate the efficiency of the animal to be scanned.

[0007] In various other embodiments, the present disclosure provides an animal retention structure, wherein the animal retention structure comprises a main chassis that is sized and structured to receive and enclose a live animal to be scanned. The animal retention structure additionally comprises an animal head yoke. The animal head yoke comprises a first yoke gate and a second yoke gate that are moveably connected to a front end of the chassis such that the head yoke can be moved between an Open position wherein the head of an animal can be placed between the first and second yoke gates and a Closed position wherein the head of the animal is restrained between the first and second yoke gates.

[0008] In yet other embodiments, the present disclosure provides a method for in vivo measuring the maintenance energy requirements of animal, wherein the method comprises measuring a surface area of a live animal utilizing computer imaging data, and performing real-time assessment of maintenance energy expenditure of the animal utilizing the computer imaging data.

[0009] More particularly, the present disclosure generally provides an animal retention, data collection and analysis system comprising multiple red, green, blue depth (RGBD) cameras that surround a retention structure and capture 2-D and 3-D images of an animal, e.g., cattle. Software is then utilized to autonomously stitch together the images from the cameras to create a single 360° surface 3-D model of the animal.

[0010] This approach is unique as it copes with the motion of the animal while the system is acquiring the images. Furthermore, the present disclosure utilizes 3-D data to predict various traits and is unique as the main parameters of interest are the surface area and volume of the animal. With the surface area, a very accurate measure of the animals basal metabolic rate (i.e., The Surface Law) can be determined. This Surface Law of animal metabolism is 25957884.v1 - 2 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 applicable from very small mice to large elephants. However, when the Surface Law was proposed, measurements of the volume of animals were not routinely collected, and the surface area and volume of animals are more predictive of basal metabolic rate.

[0011] A principal aim of this disclosure is to provide systems and method for collecting three-dimensional images in effort to more accurately measure individual differences in efficiency in cattle with computer imaging. These data will develop a greater understanding of the relationship between body volume, surface area and maintenance energy requirements of cattle that could allow for non-invasive, high-throughput phenotyping of cattle. Importantly, these data could allow for repeated, real-time assessment of maintenance energy expenditure between individual animals that could allow for animals to be sorted and managed under separate management structures designed to optimize production.

[0012] In the present disclosure the term maintenance energy requirements refers to energy used to sustain life processes of an animal in the fasting and resting state. Maintenance energy requirements in growing or lactating animals can be interpreted as the “support” cost of production and therein contribute to the inefficiency of energy use from food for production. By definition, fasting heat production is equivalent to an animal’s maintenance energy requirements; however, fasting heat production only reflects a component of those variables that contribute to metabolizable energy requirements for maintenance (MEm). Estimate of MEm are impacted by diet nutrient composition and do not closely reflect energy available to support maintenance. Correspondingly, it is generally accepted that net energy is the ideal basis for expression of energy requirements of livestock. Additionally, the present disclosure measures the maintenance energy requirements of animals at three different stages of growth of the animal (e.g., small, medium and large or ranging in size from 220 kg to 700 kg).

[0013] The present disclosure provides systems and method for the precise measurement of animal (e.g., cattle) surface area, and in various instances volume, that is then used to provide precise management of the animal’s feeding and genetic prediction of metabolic rate. 25957884.v1 - 3 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0014] This summary is provided merely for purposes of summarizing various example embodiments of the present disclosure to provide a basic understanding of various aspects of the teachings herein. Various embodiments, aspects, and advantages will become apparent from the following detailed description taken in conjunction with the accompanying drawings which illustrate, by way of example, the principles of the described embodiments. Accordingly, it should be understood that the description and specific examples set forth herein are intended for purposes of illustration only and are not intended to limit the scope of the present teachings. DRAWINGS

[0015] The drawings described herein are for illustration purposes only and are not intended to limit the scope of the present teachings in any way.

[0016] Figure 1 is an exemplary block diagram of a portable animal retention, data collection and analysis system of the present disclosure.

[0017] Figure 2 is an exemplary diagram of the portable animal retention, data collection and analysis system shown in Figure 1, in accordance with various embodiments of the present disclosure.

[0018] Figure 3 is an exemplary illustration of a 4D-RGB model generated using the portable animal retention, data collection and analysis system shown in Figure 1.

[0019] Figure 4 is an exemplary of overview of a software pipeline implemented via execution of one or more animal efficiency algorithm of the portable animal retention, data collection and analysis system shown in Figure 1, in accordance with various embodiments of the present disclosure.

[0020] Figure 5 is an exemplary multi-view point cloud registration generated by the software pipeline shown in Figure 4, in accordance with various embodiments of the present disclosure.

[0021] Figure 6 is an exemplary comparison of 3D point cloud capture quality with and without synchronization using a large box with known dimensions, the 3D point cloud capture obtained during experimentation to illustrate the 25957884.v1 - 4 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 accuracy and efficiency of the portable animal retention, data collection and analysis system shown in Figure 1.

[0022] Figure 7 exemplarily illustrates results of scanning a cylindrical object in multiple orientations, highlighting the accuracy across diverse poses of the portable animal retention, data collection and analysis system shown in Figure 1.

[0023] Figure 8 exemplarily illustrates regression analysis of predicted versus known surface area and volume for multiple static objects. The plots display the correlation between values predicted via the portable animal retention, data collection and analysis system shown in Figure 1 and the actual measurements for a cylinder, small box, medium box, and large box, all placed in the same pose across ten consecutive scans.

[0024] Figure 9 exemplarily illustrates the performance of the portable animal retention, data collection and analysis system shown in Figure 1 under direct sunlight, using a standard box to simulate outdoor livestock scanning conditions.

[0025] Figure 10 exemplarily illustrates the segmentation of cattle using combined RGB and depth models via Mask R-CNN during execution of the software pipeline during operation of the portable animal retention, data collection and analysis system shown in Figure 1.

[0026] Figure 11 is an exemplary diagram of the portable animal retention, data collection and analysis system shown in Figure 1 configured to function and operate wirelessly, in accordance with various embodiments of the present disclosure.

[0027] Figure 12 is an exemplary isometric view of an animal retention structure of the portable animal retention, data collection and analysis system shown in Figure 1, in accordance with various embodiments of the present disclosure.

[0028] Figure 13 is a top view of the animal retention structure shown in Figure 12 configured to have a head yoke in an Open position and opposing side fences in a Closed position, in accordance with various embodiments of the present disclosure.

[0029] Figure 14 is a top view of the animal retention structure shown in Figure 12 configured to have the head yoke in an Closed position and the opposing 25957884.v1 - 5 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 side fences in an Open position, in accordance with various embodiments of the present disclosure.

[0030] Figure 15 is a top view of the animal retention structure shown in Figure 12 configured to have the head yoke in a Closed position and the opposing side fences in an Open position, and having a plurality of data acquisition devices positioned adjacent the Open side fences, in accordance with various embodiments of the present disclosure.

[0031] Figure 16 is an exemplary side view of the animal retention structure absent side gates and wherein the sidewalls of the animal retention structure have large windows, in accordance with various embodiments of the present disclosure.

[0032] Corresponding reference numerals indicate corresponding parts throughout the several views of drawings. DETAILED DESCRIPTION

[0033] The following description is merely exemplary in nature and is in no way intended to limit the present teachings, application, or uses. Throughout this specification, like reference numerals will be used to refer to like elements. Additionally, the embodiments disclosed below are not intended to be exhaustive or to limit the invention to the precise forms disclosed in the following detailed description. Rather, the embodiments are chosen and described so that others skilled in the art can utilize their teachings. As well, it should be understood that the drawings are intended to illustrate and plainly disclose presently envisioned embodiments to one of skill in the art, but are not intended to be manufacturing level drawings or renditions of final products and may include simplified conceptual views to facilitate understanding or explanation. As well, the relative size and arrangement of the components may differ from that shown and still operate within the spirit of the invention.

[0034] As used herein, the word "exemplary" or "illustrative" means "serving as an example, instance, or illustration." Any implementation described herein as "exemplary" or "illustrative" is not necessarily to be construed as preferred or advantageous over other implementations. All of the implementations described below 25957884.v1 - 6 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 are exemplary implementations provided to enable persons skilled in the art to practice the disclosure and are not intended to limit the scope of the appended claims.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting. As used herein, the singular forms "a”, "an”, and "the" may be intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises", "comprising", “including”, and “having” are inclusive and therefore specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order discussed or illustrated, unless specifically identified as an order of performance. It is also to be understood that additional or alternative steps can be employed.

[0036] As used herein the phrase “operably connected to” will be understood to mean two are more elements, objects, devices, apparatuses, components, etc., that are directly or indirectly connected to each other in an operational and / or cooperative manner such that operation or function of at least one of the elements, objects, devices, apparatuses, components, etc., imparts or causes operation or function of at least one other of the elements, objects, devices, apparatuses, components, etc. Such imparting or causing of operation or function can be unilateral or bilateral.

[0037] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items. For example, A and / or B includes A alone, or B alone, or both A and B.

[0038] Although the terms first, second, third, etc. can be used herein to describe various elements, objects, devices, apparatuses, components, regions or sections, etc., these elements, objects, devices, apparatuses, components, regions or sections, etc., should not be limited by these terms. These terms may be used only to distinguish one element, object, device, apparatus, component, region or section, etc., 25957884.v1 - 7 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 from another element, object, device, apparatus, component, region or section, etc., and do not necessarily imply a sequence or order unless clearly indicated by the context.

[0039] Moreover, it will be understood that various directions such as "upper", "lower", "bottom", "top", "left", "right", "first", "second" and so forth are made only with respect to explanation in conjunction with the drawings, and that components may be oriented differently, for instance, during transportation and manufacturing as well as operation. Because many varying and different embodiments may be made within the scope of the concept(s) taught herein, and because many modifications may be made in the embodiments described herein, it is to be understood that the details herein are to be interpreted as illustrative and non-limiting.

[0040] The apparatuses / systems and methods described herein can be implemented at least in part by one or more computer program products comprising one or more non-transitory, tangible, computer-readable mediums storing computer programs with instructions that may be performed by one or more processors. The computer programs may include processor executable instructions and / or instructions that may be translated or otherwise interpreted by a processor such that the processor may perform the instructions. The computer programs can also include stored data. Non-limiting examples of the non-transitory, tangible, computer readable medium are nonvolatile memory, magnetic storage, and optical storage.

[0041] As used herein, the term module can refer to, be part of, or include an application specific integrated circuit (ASIC); an electronic circuit; a combinational logic circuit; a field programmable gate array (FPGA); a processor (shared, dedicated, or group) that performs instructions included in code, including for example, execution of executable code instructions and / or interpretation / translation of uncompiled code; other suitable hardware components that provide the described functionality; or a combination of some or all of the above, such as in a system-on-chip. The term module can include memory (shared, dedicated, or group) that stores code executed by the processor.

[0042] The term code, as used herein, can include software, firmware, and / or microcode, and can refer to one or more programs, routines, functions, classes, and / or objects. The term shared, as used herein, means that some or all 25957884.v1 - 8 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 code from multiple modules can be executed using a single (shared) processor. In addition, some or all code from multiple modules can be stored by a single (shared) memory. The term group, as used above, means that some or all code from a single module can be executed using a group of processors. In addition, some or all code from a single module can be stored using a group of memories.

[0043] Although the systems and methods of the present disclosure are applicable to any type and size of live animal, for simplicity the systems and methods of the present disclosure will be exemplarily described and illustrated herein with regard to live cattle.

[0044] Recent advances in computer vision and image processing have created exciting opportunities to directly measure factors that control the efficiency with which cattle use nutrients to produce meat or milk. And these advances in computer vision can now provide a tool to rapidly phenotype efficiency in cattle. A goal of the present disclosure is to utilize computer vision to provide rapid and cost-effective techniques that can accurately identify individual differences in the efficiency with which cattle use nutrients for growth or lactation in real-time. Development of real-time measures of individual differences in the efficient use of nutrients by cattle will allow for implementation of novel management techniques that will provide large benefits in animal productivity, economic sustainability, and also reduce environmental impacts of cattle production systems. Additionally, development of tools that can repeatedly collect accurate measures of different phenotypes in cattle will improve accuracy of genetic tools designed to identify superior animals for breeding and allow novel insights about an animal’s genetic potential for efficiency together with influences of diet and other environmental factors.

[0045] Improving the efficiency with which animals (e.g., ruminants, e.g., cattle) use feed is needed to produce greater amounts of food for human nourishment, and a number of efforts have attempted to develop tools to identify more efficient animals. Yet, realizing broad reaching improvements in efficiency in production of such things as beef and milk is difficult because the efficiency complex is multifaceted and involves aspects of physiology, genetics, environment and economics. Individual differences in efficiency correspond to differences in 25957884.v1 - 9 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 digestion and metabolism that alter the net energy available from feed. However, individual measurements of digestion and metabolism are costly and difficult to determine across large populations. Thus, most tools designed to identify more efficient animals have focused on estimating an animal’s genetic merit for efficiency, in order to identify superior animals for use in breeding.

[0046] Unfortunately, genetic merit explains less than 50% of the individual differences in efficiency in cattle (estimates range from approximately 20% to 50%) and measurements of efficiency in growing cattle are only marginally related to measurements of efficiency in mature cattle. Limitations in the ability to identify individual differences in efficiency from genetic tools alone is perhaps not surprising, because nearly all genetic investigations in efficiency have been based on measures of complex biological processes (e.g., growth and intake) and have not focused on the principal factors (e.g., metabolism, energy expenditure) that contribute to efficiency. Further, these genetic predictions have been limited by the use of body weight to measure metabolic needs, rather than surface area and volume which are directly related to the thermodynamics of maintenance energy requirements. The current inability to identify individual differences in efficiency in cattle prevent large improvements in production, economic returns and environmental sustainability because animals are unable to be effectively matched with optimal management systems. Thus, there is a critical need to accurately identify efficient individuals in real-time. Without the ability to effectively identify efficient cattle then matching management strategies to individual cattle to improve the efficiency will be difficult. As a consequence, large advancements in productivity, economic viability and environmental sustainability of food production from cattle will remain elusive.

[0047] Generally, the present disclosure provides a portable animal retention, data collection and analysis system 10 that comprises an animal retention structure 14 structured and operable to temporarily restrain an animal (e.g., a cow) in specific location, an animal data acquisition system 18 structured and operable to acquire certain data of the animal (e.g., surface area, surface temperature and / or volume), and a computer-based data processing system 22 that is structured and operable implement and execute one or more animal efficiency algorithm (e.g., 25957884.v1 - 10 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 computer software, programs, and / or code) that utilizes the acquired data to calculate the efficiency of various animals (e.g., cattle) and then measure individual differences in efficiency in the various animals. An objective of the present disclosure is to develop a greater understanding of effects of various genetic and environmental factors on efficiency in cattle. The 3D imaging data will allow for accurate and precise prediction of individual differences in efficiency and maintenance energy requirements of cattle in real-time. The 3D imaging techniques disclosed herein will allow high throughput phenotyping of cattle and have a transformative effect on the understanding of factors that influence efficiency in ruminants. A deeper understanding of factors that influence efficiency in ruminants will aid in discovering why individual differences in efficiency among cattle change in response to time, diet and environment. Furthermore, the repeated measures of maintenance energy requirements will help to improve productivity by enhancing identification of morbid cattle, improving animal well-being and reduce overuse of antibiotics in cattle production systems by more effectively deploying treatments to sick animals at the onset of an immune response. As used herein the word portable will be understood to mean that all components, devices, systems, structures, mechanisms, modules, apparatus, etc. of system 10 can be easily assembled and disassemble and loaded into a vehicle (e.g., a truck, SUV or van) and / or onto a trailer that is towable by the vehicle and readily transported from one location to another.

[0048] The efficiency of an animal (e.g., cattle) is limited by the ability to use energy for productive purposes. Efficiency can be defined in many ways; in livestock, efficiency is commonly calculated as the amount of product (e.g., gain in body mass, milk produced) for a unit of input (e.g., amount of feed or nutrient consumed). Livestock support processes are important to life by breaking down feed to its simplest components and then organizing the components into substances that support life. Thus, livestock grow, lactate and sustain themselves by creating local increases in organization of simple components derived from digested feed; yet, physical laws require increases in entropy, not increases in organization. Consequently, productive processes of livestock stand in opposition to an increasingly chaotic environment by creating local increases in organization, 25957884.v1 - 11 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 which requires a persistent expenditure of energy to sustain life and to be productive. The amount of effort expended by livestock to create greater organization from an increasingly disordered environment is not inconsequential, and can account for one-third to more than all of the daily energy consumed by an animal. Hence, it becomes fairly apparent how growth, lactation and survival (i.e., maintenance) are most limited by energy available to animals.

[0049] Historically measures of basal energy requirements have been intensive. Energy itself is an abstraction and defined as the ability to do work. Perhaps the most salient features of energy, however, are that it is neither created nor destroyed and that energy losses from a body are emitted as either heat or light. Cattle and other non-genetically engineered livestock animals are not bioluminescent and are unable to lose energy absorbed from nutrients as light. Therefore, energy losses from nutrients absorbed by cattle are transferred through heat losses to their environment. Cattle and other homeotherms must closely regulate their internal body temperature to support life’s processes. The internal body temperature of cattle (38.6 °C / 101.48 °F) is often greater than that of their environment. Thus, heat losses are a direct function of the surface area, and in various instances volume, of cattle and surface area:volume ratio. This relationship is well known, and because it is conserved across a large number of homeotherms many authors have referred to the relationship between body surface area and maintenance energy requirements as the “Surface Law”. However, prior to the present disclosure, the relationship between heat loss, surface area and volume has not been studied

[0050] Measurements of maintenance energy requirements are relatively straightforward but are intensive to collect. Since energy is neither created nor destroyed the maintenance energy requirements can be measured by determining heat losses of an animal that is fasted. During periods of fasting an animal survives in a state where no energy inputs are derived from food. Accordingly, energy losses during fasting represent the amount of energy expended to support survival. Energy losses from animals can be determined directly by measuring the transfer of heat from animals. However, in the past accurately measuring the direct transfer of heat 25957884.v1 - 12 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 from an animal would require a highly controlled environment and sensitive equipment. Thus, direct measures of energy losses from animals rapidly has given way to measurements of energy retained in the body, which allows for calculation of energy losses by difference. However, measurements of energy retained in the body typically require measures of the energy content in the body, which previously has been evaluated by complete combustion of samples obtained from all body tissues. Such direct measures of energy retained in the body are destructive, because they require the combustion of the animal of interest. Clearly, combustion of animals shortly after entry in a production system to determine their potential for efficient production of meat or milk is deleterious to future productivity and economic sustainability of livestock production systems.

[0051] Recent attempts to facilitate repeated measures of energy retention by non-destructive measures using bioelectrical impedance have been evaluated. However, measures of bioelectrical impedance are laborious and stressful to cattle. Useful measures of bioelectrical impedance in cattle requires the internal placement of metal probes near the top of the shoulder and tail of cattle to overcome variation in resistance associated with hair, hooves and mud. Estimation of body composition based on bioelectrical impedance has an inherently large amount of basal variation because this approach assumes that all non-adipose tissues have the same amount of impedance. Furthermore, determining efficiency of energy use with bioelectrical impedance requires multiple estimates of body composition and measures of intake.

[0052] Another commonly used technique to measuring energy losses in cattle for the past 70 years has involved calculation of heat of combustion from the respiratory exchange of oxygen, carbon dioxide and methane. This approach is relatively easy, because it can be accomplished using open-circuit calorimeters that do not require the complete exclusion of environmental air. Also, open-circuit calorimetry allows non-destructive, repeated measures of energy losses with considerably greater amounts of accuracy and precision in comparison to measures of electrical impedance; however, measures of energy losses using open-circuit calorimetry still require animals trained to the procedure. Adequate precision in measures of heat expenditure via open-circuit calorimetry often need several days 25957884.v1 - 13 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 of measurement (at least 2 days) to achieve acceptable levels of precision in measures and direct measures of maintenance energy requirements require the animal be placed in a fasting state of metabolism.

[0053] Fasting metabolic states in nonruminant animals occur relatively rapidly, but rates of fermentation and digesta passage from the rumen prevent ruminant animals from achieving a fasting metabolic state until feed has been withheld for multiple days. Therefore, technical aspects in obtaining measures of energy losses in fasting cattle via open-circuit calorimetry to identifying individual differences in efficiency prevent large-scale use of open-circuit calorimetry.

[0054] Measures of body surface area can ideally describe basal energy requirements. Differences in metabolic rates within and across various species have been of interest to investigators for over 200 years. Early investigations revealed that metabolic rate across a number of different homeotherms and body sizes were closely related. The coordinate discoveries that larger homeotherms produce more heat than smaller homeotherms and that rate of heat production expressed in equal units of body weight is less for large animals than for small animals seemed to imply that metabolic rate of animals change as animals increase in mass. However, several early investigators reported that metabolic rate was conserved when expressed as a proportion of the total surface area of the body. The concept that metabolic rate is constant as a proportion of body surface area is referred to as the ‘Surface Law’. The concept of the ‘Surface Law’ is based on the first law of thermodynamics (i.e., conservation of energy). Specifically, uniform metabolic rates as a function of body surface area in homeotherms are required to account for rates of heat transfer from a body’s surface to its surroundings.

[0055] Despite the understanding that differences in metabolic rates expressed as a proportion of body surface area are uniform across animals and that differences in metabolic rate as a function of body mass are non-uniform, considerable efforts have been made to estimate metabolic rates and energy requirements through mathematical transformation of body mass. However, an index based on body mass that can provide harmony between measures of an animal’s mass and its metabolic rate are based on the relative difficulty in measuring 25957884.v1 - 14 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 body surface area in comparison to measures of body mass. Previous measures of body surface area were tedious, imprecise and often collected using destructive measures (i.e., removing the skin from the animal). Thus, measures of body surface area are limited to use in academic discovery via post hoc analyses and were not well-suited to determination of metabolic rates of live animals.

[0056] The theory that mathematical transformation of body mass can be precisely related to metabolic rate is dependent on the principle that large and small bodies of similar shape have surface areas in proportion to the square of their linear dimensions or the two-thirds power of their volume (i.e., square-cube principle). That is, if two bodies have the same density, then their surface areas should be proportionate to the two-thirds power of their mass. Unfortunately, estimates of surface area based on measures of body mass have inherent variation that is an exponential function of inter- and intra-species differences in the relationship of body shape and body mass. Efforts to improve estimates of surface area from measures of body mass have been suggested by multiplying the power of mass by a constant value that is determined empirically from measures of surface area and mass for a given species. Unfortunately, this approach is also predicated on an ideal relationship between body surface and body mass, and therefore only useful in improving estimates of surface area from body weight when the relationship between body surface area and body mass are perfectly related across all possible body shapes and masses within a given species. The inherent flaw in this approach to estimating body surface area results in an unending set of ‘new’ constants for a myriad of different species. This approach to estimating surface area results in estimates of ‘metabolic body size’ calculated from body mass that deviates from the square-cube principle. Further, these estimates ignore the interplay of surface area and volume on metabolic rate.

[0057] A great body of work related to metabolic rate has been conducted in cattle. Current estimates of metabolic rates in cattle are generally expressed as a constant or metabolic rate multiplied by metabolic body size (Maintenance energy requirement = 70 kcal ×〖body weight, kg〗3 / 4. Ultimately, it is unclear if calculations of metabolic body size that deviate from the square-cube principle represent non-uniform heat loss across different areas of the body’s 25957884.v1 - 15 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 surface, variance in body shape of a given species at the time of measurement or changes in the uniformity of heat loss or body shape in response to evolutionary progress.

[0058] The first reported surface area measurements in animals was done by covering the body of 16 humans in sections with strips of millimeter graph paper to obtain surface area measurements. Little improvement has been made in collecting non-destructive measures of surface area in cattle since, but several recent reports have demonstrated that 2D computer imaging of the dorsal aspect of cattle can predict weight gain and body composition. Furthermore, assessment of weight gain is improved when 2D computer image analysis was combined with biometric measures of height and width. Weight gain and lactation are predominately related to efficiency of energy use in cattle. Recent discoveries in the use of computer imaging are encouraging, but these efforts have not included measures of total body surface area.

[0059] To date, the most advanced non-destructive technique described to obtain total body surface area measurements of cattle was developed as the “surface area integrator”. These techniques covered cattle with an ink roller, counted the number of revolutions and calculated surface area by multiplying the area of the roller by the number of revolutions. Clearly, approaches used by early investigators do not allow for high throughput phenotyping, and it is easy to understand why a number of the studies that contributed to the early discovery of the “Surface Law” were limited to a relatively small number of subjects. And it is understandable why efforts were subsequently made to estimate surface area from measures such as body mass that were much easier to collect. Early measures of surface area contributed to the synthesis of allometric equations to estimate surface area and later this concept served as the basis for what is now referred to as “metabolic body size” in cattle. Yet, even prior to the development of metabolic body size it was understood that there are differences in the relationship between body mass and surface area across and within animals in response to differences in age and overall body size.

[0060] In the past 90 years nearly no progress has been made in developing a more effective system to measure individual differences in 25957884.v1 - 16 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 maintenance energy requirements in cattle other than those dependent on measures of mass (i.e., metabolic body size). Inaccuracies inherent with the use of a single index (i.e., metabolic body size) to predict maintenance energy expenditures are the most probable contributors to inaccuracies in estimates of efficiency from large datasets of growing cattle fed in the United States.

[0061] Despite this lack of progress in adapting novel techniques to directly measure surface area in cattle, other fields of study have made good progress in developing new methods in use of less invasive technologies that allow for the measurement of surface area in other species. As a result of the importance of surface area in the treatment of human health a, number of successful techniques have been established to collect non-invasive direct measures of surface area of humans that provide greater accuracy than the use of a single index based on body mass. Measures of body surface area are important in the treatment of burn patients, liver transplantation and estimating renal function in the diagnosis of kidney diseases. Recent approaches to measuring surface area have included three-dimensional (3D) laser scanning in humans and use of computed tomography in humans and rabbits. Use of 3D laser scanning can provide accurate, non- destructive, repeated measures of the surface area of living animals; however, these approaches require expensive and technically complex equipment. Furthermore, these technologies often require the subject to avoid movement for a prolonged period of time.

[0062] Reconstruction of the 3D geometry of rigid objects or static environments, including the generation of point clouds, relies on computer vision concepts developed over the past 50 years. In traditional multi-stereopsis, 3D structure can be extracted from a relatively small set of partially overlapping images emanating from multiple cameras, arranged in an array of positions to cover the entire view of the target object or scene. In those cases, prior knowledge of scene geometry, camera parameters, camera pose (position and orientation), and ground control point are required. However, in more recent methods such as Simultaneous Localization and Mapping and Structure from Motion, neither ground control, reference targets, nor a priori knowledge of the camera parameters and their poses are needed. Instead, the 3D geometry of the scene and the camera extrinsic 25957884.v1 - 17 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 parameters are solved simultaneously, with little or no user intervention. These approaches originated from the robotics simultaneous localization and mapping (SLAM) and the computer vision communities structure from motion (SfM). They incorporate automatic feature-matching algorithms and a robust minimization of the error between the detected and re-projected features associated with the 3D structure (point clouds). The error minimization SfM, known as Bundle Adjustment succeeds if a large set of multiple partially overlapping images can lead to a rich number of matching features. In that case, these simultaneous, highly redundant, iterative bundle adjustment procedures together with a large database of automatically extracted features result in a very accurate point matching and hence, an extremely dense RGB-encoded point cloud (3D-RGB). A sub-product of this processing is the ability to reconstruct the sequence of motions (rotations and translations) of the cameras. Once again, it should be emphasized that the accuracy of the structure and the motion sequence obtained, as well as the density of the point clouds are highly dependent on the quality and amount of overlapping of the images while their ability to cover the full 3D structure of the object or scene is preserved.

[0063] Measurement of surface area, and various instances volume and surface area:volume ratio, can provide insights to improve intake models. All animals’ plane of nutrition is ultimately a function of the nutrient composition of the diet and the amount of diet consumed. Arguably, variations in amounts of diet consumed contribute most to various states of malnourishment in comparison to consumption of diets that lack an ideal nutrient composition. Thus, developing a greater understanding of factors that influence intake of food by animals is important to understanding factors that impair growth, health and reproduction. When ruminants consume nutrient-dense diets, short-term regulation of food intake in ruminants is controlled by oxidation of absorbed nutrients, and long-term regulation of food intake is predominately controlled through a complex array of humoral signals. However, perhaps the greatest ecological niche of ruminants is utilization of plant-based materials indigestible to mammalian enzymes for nourishment. Mechanisms controlling food intake in ruminants largely differ from nonruminants when ruminants consume diets predominately comprised of plant fibers. Slow rates 25957884.v1 - 18 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 of fiber fermentation in the rumen can contribute to physical fill of the ruminal compartment, and pressure sensitive ganglion signal satiety when ruminants graze slowly fermentable forage. Inter- and intra-species ruminal volume is generally theorized to be a function of body size, which has historically been measured by body mass. It seems likely that ruminal volume is more closely related to body volume in comparison to body mass. Thus, development of computer imaging technologies that allow repeated, real-time, non-destructive measures of body volume could facilitate more accurate predictions of intake among the majority of cattle managed in extensive grazing systems; where direct determination of intake is difficult.

[0064] Feed costs represent a majority of the production costs of meat and milk from cattle in the United States. Improving feed use efficiency ultimately represents one of the greatest targets to reduce production costs. The efficiency with which cattle use energy derived from diet to support growth or lactation is directly related to the efficiency with which cattle use feed for production because energy first-limits the ability of cattle to grow and lactate.

[0065] All energy derived from metabolism of nutrients absorbed by cattle is used to support the fundamental processes needed for survival (i.e., maintenance) and biological fitness (i.e., growth, lactation or reproduction). Energy used for maintenance is most important to sustaining life, and energy absorbed from nutrients are first partitioned to maintenance processes prior to being used to support the higher order processes of growth, lactation or reproduction. Energy used for maintenance is not retained in the body. Thus, energy from feed used to support processes essential to survival represent an inefficiency in the use of feed for growth or lactation.

[0066] There are large individual differences in the efficiency with which cattle use feed for productive purposes. An experiment measured differences in feed efficiency across nearly 4,000 beef cattle and reported that difference between the maximum and minimum amount of feed used by large groups of cattle for growth could be as large as 12 lbs on a dry matter-basis (i.e., 4-times the phenotypic standard deviation). Using current feed costs and days on feed of growing cattle fed in confinement for slaughter, a 12 lb difference in dry matter 25957884.v1 - 19 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 intake is equal to $281 per animal. If individual differences in efficiency of feed use for growth is equally distributed across the nearly 26 million cattle fed in confinement for slaughter in the United States at not more than 1 standard deviation from mean required 12 lbs greater intake for a pound of growth then the cost of less efficient cattle currently adds over $1.2 billion in annual feed costs to feedlot production systems alone. If less efficient cattle in feedlots are also less efficient in grazing systems then these costs would be even greater.

[0067] Inefficient cattle fed forage in extensive grazing systems are nearly impossible to identify due to inability to measure feed consumption in extensive environments. Inefficient cattle in these systems contribute to reductions in efficiency of use of land and forage resources and increase production of environmentally reactive gasses from cattle production systems. Forage fed cattle account for over 90% of methane contributions from cattle. Methane is an environmentally reactive gas produced in the rumen and represents energy losses from feed. Thus, a system that provides high throughput phenotyping of cattle in extensive grazing systems, such as the portable animal retention and data collection system described herein could be used in the identification of cattle likely to reduce economic returns, and cattle that provide greater contributions of environmentally harmful gases.

[0068] Grazing cows account for nearly 75% of the overall energy requirement for beef production, and inefficient cows are much more likely to have a greater impact on long-term costs and environmental impacts of production, because these animals experience a greater life cycle than animals used for meat production. Additionally, inefficient cows have the potential to continue to contribute inefficient genes to future offspring. It is also difficult to identify individual differences in efficiency with which cattle use feed for lactation, because cows will often mobilize body lipids to support lactation even when caloric intake from the diet is limited. Thus, cows that acquire less calories from feed to support lactation can appear efficient if they can adequately mobilize lipid stores to support their genetic potential for lactation. Development of a method that can accurately and repeatedly evaluate individual differences in efficiency of cows, such as the methods implemented utilizing the portable animal retention and data collection system 25957884.v1 - 20 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 described herein, could therefore have even greater impacts on the economic and environmental sustainability of production of meat and milk from cattle.

[0069] Measurements of maintenance energy requirements can allow for determination of individual differences in efficiency even though a large number of complex biological and environmental factors influence the efficiency with which cattle use feed for growth or lactation. Identifying cattle with greater relative maintenance energy requirements could allow development of low-input management strategies to optimize economic returns from these animals. For example, less efficient cattle could be fed less energy dense diets that are also less costly. Also, systems that facilitate the repeated non-destructive measures of maintenance energy requirements would help to better select animals with superior genetic potential for breeding and provide long-term advancement in the efficiency of cattle production systems.

[0070] Maintenance energy requirements represent the amount of energy needed to maintain the body at a given moment in time. As animals grow and experience different environmental conditions, maintenance energy requirements change. Indeed, a primary justification to the use of metabolic body size rather than measures of surface area to identify maintenance energy requirements in cattle was that the surface area of a living animal is not constant and cannot be measured in such a manner that the results can be checked by different investigators. Although, some have argued that fasting heat production overestimates maintenance energy requirements, because fasting heat production includes heat production associated with generating adenosine triphosphate (ATP) from body nutrients in addition to maintenance energy requirements. This is an interesting argument, even though it ignores previous definitions of maintenance energy requirements. However, data collected utilizing the system and methods of the present disclosure are envisioned to be instrumental in furthering knowledge of the relationship between fasting heat production, surface area, volume and maintenance energy requirements, because data generated from the portable animal retention and data collection system described herein includes measures of energy intake and direct measures of energy retained in the body. Furthermore, even if fasting heat production should be corrected for efficiency of converting body 25957884.v1 - 21 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 nutrients to ATP during fasting (kb) in the calculation of maintenance energy costs, measures of surface area and volume should still be more closely related to maintenance energy expenditure in comparison to estimates based on mass, because kb likely has little variation between animals and can be included to correct linear models.

[0071] Considering the complexity and labor required to collect measures of surface area in animals it is understandable why early investigators were willing to adopt the metabolic body size system despite inherent flaws that prevent accurate identification of individual differences in efficiency. However, advances in computer imaging technology make it easier to acquire rapid and accurate non-invasive measures of the surface area of cattle without complex and costly equipment. Use of computer imaging technologies allow direct use of surface area to determine maintenance energy requirements rather than an index based on body mass. Ultimately, the ability to evaluate individual differences in maintenance energy requirements aid the development of optimal management strategies that pair an animal’s potential for efficient production of meat and milk with appropriate inputs. Of equal or even greater importance is the ability to rapidly enhance genomic prediction models that will help to pursue the most efficient cow.

[0072] Referring now to Figure 1, the present disclosure provides systems and methods for accurate and precise in vivo measurement of individual differences in efficiency with which cattle use nutrients for growth or lactation in real- time utilizing measured surface area, surface temperature, and / or volume of an animal, thereby providing high throughput phenotyping of maintenance energy requirements in real-time. More particularly, the present disclosure provides the portable animal retention, data collection and analysis system 10 and methods for analyzing the data collected by the system 10 to enable repeated, real-time assessment of maintenance energy expenditure between individual live animals that will allow for animals to be sorted and managed under separate management structures designed to optimize production. For simplicity, the portable animal retention, data collection and analysis system 10 be sometimes referred to herein simply as the system 10. 25957884.v1 - 22 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0073] As described above, the system 10 comprises the portable animal retention device 14, the portable animal data acquisition system 18 and the portable computer-based data processing system 22. The animal retention structure 14 can be any device, apparatus, system or structure suitable for temporarily restraining an animal in a target location such that various phenotypic data of the animal (e.g., surface area, surface temperature and / or volume) can be acquired by the data acquisition 18. For example, in various embodiments, the animal retention structure 14 comprises a portable chute, corral, pen or cage having a gate or door that can be opened to allow the animal to be positioned within the chute, corral, pen or cage, and then closed to restrain the animal within the chute, corral, pen or cage. The animal data acquisition system 18 can be any system comprising one or more data acquisition device 24 structured and operable to acquire phenotypic data of the animal restrained withing the animal retention structure 14. For example, the data acquisition device(s) 24 can comprise one or more red-green-blue (RGB) cameras, and / or one or more charge-coupled device (CCD) camera, and / or one or more time-of-flight sensor, and / or one or more infrared (IR) camera, and / or one or more thermal sensor / camera, and / or one or more X-ray device, and / or one or more magnetic resonance imaging (MRI) camera, etc. that is / are structured and operable to acquire various desired phenotypic data of the animal temporarily restrained within the retention structure 14. The data acquisition devices 24 can be located in any location near or adjacent to the retention structure 14 such that the data acquisition devices 24 have a desired field of view or line of view and are free from disturbance by the animal restrained within the animal retention structure 14. For example, in various embodiments, one or more data acquisition device 24 can be mounted to a respective adjustable tripod position outside at perimeter of, and free from contact with, the animal retention structure 14, such that each data acquisition device 24 can be located, oriented and positioned to have predetermined desired field of view or line of view, and not subject to and removed from any possible disturbance caused by movement of the animal within the animal retention structure 14.

[0074] In various embodiments, the data acquisition devices 24 of the data acquisition system 18 can comprise a plurality of digital RGB cameras 25957884.v1 - 23 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 positioned around the animal retention structure 14 to acquire three-dimensional (3D) digital image data of the animal temporarily restrained within the retention structure 14 from multiple viewing angles and perspectives, a plurality of time-of- flight (ToF) sensors or cameras positioned around the animal retention structure 14 to acquire ToF data of the animal temporarily restrained within the retention structure 14 from multiple viewing angles and perspectives, and in some instances one or more thermal sensor or camera structured and operable to obtain thermal data from the animal temporarily restrained within the retention structure 14. The data processing system 22 is communicatively connected (wired and / or wirelessly) to the data acquisition system 18 (i.e., to the components thereof) and can be any computer-based device or system that is structured and operable implement and execute one or more animal efficiency algorithm (e.g., computer software, programs, and / or code) that receive as input the data output from the data acquisition system 18 (i.e., the components thereof) and utilize the received data to determine individual differences in efficiency in various animals that have been processed through the system 10.

[0075] In various embodiments, the system 10 is structured and operable to measure surface area, and / or volume of individual live cattle, and in various instances thermal data, and generate data related to the relationship of factors (i.e., diet, age, environment, genetic merit) that influence the efficiency with which cattle use nutrients for growth or lactation in real-time (referred to herein simply as efficiency). Moreover, the system 10 provides a computer vision-based system that can be deployed to identify individual differences in efficiency among a certain type of animal (e.g., among cattle) that can be broadly used to formulate optimal management and breeding decisions. Furthermore, the system 10 is structured and operable to simultaneously provide 1) a model to predict maintenance energy requirements based on computer vision of individual measures of surface temperature, and / or surface area, and / or volume together with direct observations of individual maintenance energy requirements and measures of genetic merit for feed efficiency, 2) utilize computer vision to better identify the relationship between surface area, feed efficiency and feed intake among cattle fed forage- or grain- based diets, and 3) identify genetic parameters, genomic associations and genetic 25957884.v1 - 24 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 correlations (including maintenance energy) from four-dimensional (4D) image data, and identify relevant production traits to develop novel genetic predictions that will serve as an available tool in making improved management and breeding decisions for cattle. As used herein 4D image data will be understood to mean the combination of XYZ point could (3D) plus the corresponding temperature at that same XYZ coordinate. Similarly, 3D-RGB images consist of the same XYZ point cloud (3D) plus the so-called texture or colors (RGB) at that same XYZ coordinate

[0076] Referring now to Figures 1 and 2, as described above, in various embodiments, the data acquisition system 18 comprises a plurality of data acquisition devices 24 that are positioned near or adjacent to the retention structure 14 such that they are isolated from the animals and free from disturbances due to the animals’ movements. Furthermore, the data acquisition devices 24 are positioned such that their combined fields of view or lines of sight create a volume large enough to cover the entire animal positioned within the retention structure 14.

[0077] In various instances the data acquisition devices 24 can comprise a plurality of RGB cameras 26, and a plurality of infrared time-of-flight (ToF) cameras / sensors 30, and in some instances a plurality of thermal sensors / cameras 34. Additionally, in various embodiments, the data processing system 22 comprises a data acquisition system control center 38 and a computer-based system controller 42 that is communicatively connected (wired or wirelessly) to the data acquisition system control center 38. The data acquisition device control center 38 is communicatively connected (wired or wirelessly) to each of the data acquisition devices 24 and comprises a plurality of data acquisition device control modules 46 that are each structured and operable to control the operation (e.g., data acquisition and data transmittal) of one or more of the data acquisition devices 24. It is envisioned that the data acquisition device control modules 46 can be hardware and / or software-based modules that are structured and operable to control the operation (e.g., data acquisition and data transmittal) of one or more of the data acquisition devices 24 as described herein. For example, it is envisioned that the data acquisition device control modules 46 can comprise one or more application specific integrated circuit (e.g., ASIC(s)), combinational logic circuit(s); field programmable gate array(s) (FPGA); processor(s) (shared, dedicated, or group) 25957884.v1 - 25 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 that execute software code; and / or other suitable hardware components that provide the functionality described herein. The computer-based system controller 42 can be any one or more computer-based component, device or system (e.g., one or more laptop computer, desktop computer, tablet, dedicated computer node, or combination thereof) that comprises one or more processor and one or more electronic data storage device that contains a computer-readable medium (e.g., a hard drive). Generally, the electronic data storage device can be any device suitable for storing such things as the various software, programs, algorithms, code, digital information, data, look-up tables, spreadsheets and / or databases, etc., used and executed during operation of the system 10, as described herein.

[0078] In various embodiments, the plurality of RGB cameras 26 can comprise eight RGB cameras 26, each structured and operable, as controlled by a respective one of the data acquisition device control modules 46, to acquire digital image data of the live animal and transmit the acquired image data to the system controller 42. For example, the eight RGB cameras 26 can include a left head RGB camera 26, a right head RGB camera 26, a left-side bottom RGB camera 26, a left- side top RGB camera 26, a right-side bottom RGB camera 26, a right-side top RGB camera 26, a front RGB camera 26 and a rear RGB camera 26. In such instances, the left head RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the left side of the animal’s head, and right head RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the right side of the animal’s head such that their combined fields of view encompasses the head and neck of the animal. The left-side top RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the upper and top portions of the left side of the animal’s torso, and the right-side top RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the upper and top portions of the right side of the animal’s torso, such that their combined fields of view encompasses the superior torso of the animal. Similarly, the left-side bottom RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the bottom and lower portions of the left side of the animal’s torso, and the right-side 25957884.v1 - 26 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 bottom RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the bottom and lower portions of the right side of the animal’s torso, such that their combined fields of view encompasses the sides and inferior portions of the animal. The front RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the front end of the animal’s torso, and rear RGB camera 26 is positioned near or adjacent to the retention structure 14 to have a field of view that encompasses the rear end of the animal’s torso.

[0079] In various embodiments, the ToF sensors 30 comprise infrared sensors that are structured and operable to determine the distance between the respective ToF sensor 30 and the animal. In various instances, the plurality of time- of-flight sensors 30 can comprise eight time-of-flight sensors 30, each structured and operable, as controlled by a respective one of the data acquisition device control modules 46, to acquire time-of-flight data (i.e., distance data) and transmit the acquired ToF data to the system controller 42. The ToF sensors 30 are mounted to or are integrally formed with or otherwise located in close proximity to the RGB cameras 26. The ToF data is acquired simultaneously with the RGB digital data and is utilized to determine the distance between the ToF sensor 30 (more particularly, RGB camera 26) and the animal when each set of RGB image data is acquired.

[0080] The distance information is combined with the RGB image data and is utilized by the animal efficiency algorithms to more accurately determine surface area, and in various embodiments the volume and surface area:volume ratio, of the animal, and thereby more accurately predict maintenance energy requirements as provided by the more accurately determined surface area, as described in detail below.

[0081] In various instances, the eight time-of-flight sensors 30 can include a left head time-of-flight sensor 30, a right head time-of-flight sensor 30, a left-side bottom time-of-flight sensor 30, a left-side top time-of-flight sensor 30, a right-side bottom time-of-flight sensor 30, a right-side top time-of-flight sensor 30, a front time-of-flight sensor 30 and a rear time-of-flight sensor 30. In such instances, the left head time-of-flight sensor 30 is positioned near or adjacent to the retention 25957884.v1 - 27 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 structure 14 to have a line of sight or field of view directed at the left side of the animal’s head, and right head time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the right side of the animal’s head, such that their combined lines of sight or fields of view encompasses the head and neck of the animal. The left-side top time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the upper and top portions of the left side of the animal’s torso, and the right-side top time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the upper and top portions of the left side of the animal’s torso, such that their combined fields of view encompasses the superior torso of the animal. Similarly, the left-side bottom time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the bottom and lower portions of the left side of the animal’s torso, and the right-side bottom time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the bottom and lower portions of the right side of the animal’s torso such that their combined fields of view encompasses the sides and inferior portions of the animal. The front time-of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the front end of the animal’s torso, and rear time- of-flight sensor 30 is positioned near or adjacent to the retention structure 14 to have a line of sight or field of view directed at the rear end of the animal’s torso. Generally, as described further below, the system 10 utilizes the 3D RGB image data acquired by RGB cameras 26 and the ToF data acquired by the ToF sensors 30 to generate automatic point cloud registrations. This automatic point cloud registration allows for flexible setup and positioning of the data acquisition devices 24 that is not tied to any specific arrangement.

[0082] As described above, in various instances the data acquisition devices 24 include a plurality of thermal sensors 34 that can comprise at least one thermal sensor 34 positioned near or adjacent to the retention structure 14 to have temperature sensing field directed at a rearward portion of the animal’s torso, and at least one thermal sensor 34 positioned near or adjacent to the retention structure 25957884.v1 - 28 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 14 to have temperature sensing field directed at a forward portion of the animal’s torso. In such instances, the temperature data can be combined with the distance information and the RGB image data and is utilized by the animal efficiency algorithms to more accurately determine heat loss of the animal, and thereby more accurately predict maintenance energy requirements as provided by the more accurately determined surface area, and in various instances volume, as described in detail below.

[0083] In operation, once a live animal is restrained within the animal retention structure 14, as described further below, an operator initiates a data capture process, via the data processing system 22, i.e., via the system controller 42, to capture the RGB image data and the ToF data, and in various instances thermal data. Particularly, the data is capture via the RGB cameras 26 and the ToF sensors 30, and in various instances the thermal sensors 34, which are controlled (wired or wirelessly) by the data acquisition device control modules 46, which are controlled (wired or wirelessly) by the system controller 24 via execution of the one or more animal efficiency algorithm(s). The data is then transmitted (wired or wirelessly) to the system controller 42 where the data is utilized to generate, via execution of the one or more animal efficiency algorithm(s), 3D-RGB models, or in instances wherein the thermal sensors 34 are incorporated in the system 104D- RGB models (i.e., 3D RGB image data + thermal data), and therefrom compute the animal’s surface area and volume, as exemplarily illustrated in Figure 3.

[0084] It is envisioned that in various instances the animal can be tagged with an electronic ID tag whereby once a live animal is restrained within the animal retention structure 14 the animal’s ID tag can be read by an electronic identifier (EID) or radio frequency identifier (RFID) and the process can be automatically initiated based on the presence the electronic identifier (EID) or radio frequency identifier (RFID).

[0085] Advanced computer vision methods are implemented, via execution of the animal efficiency algorithm(s), to autonomously estimate surface area and volume from the 4D-RGB data. In surface from motion (SfM), the quality and the amount of overlapping between images affect the quality of the 3D point cloud, and in consequence the derived phenotypes, such as surface area and 25957884.v1 - 29 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 volume of the target animal. It is known that typically high-dense and accurate 3D reconstruction of objects require a very large number of cameras. This problem can be exacerbated by the lack of texture of the skin of animals. A smooth, high- curvature, and texture-poor surface does not generate enough feature or key points to be matched between overlapping images (using SIFT, SURF, ORB, etc.). However, the system 10, as described herein, utilizes a significantly smaller number of cameras and sensors (e.g., the RGB cameras 26 and / or ToF sensors 30) without compromising density and accuracy of the 3D models created. Particularly, via execution of the animal efficiency algorithm(s) the reliance on matching of 2D key points is reduced, and instead utilized the matching of the re-projected 3D points onto the image plane by introducing the combination of RGB cameras 26 and the ToF sensors 30. Moreover, via execution of the animal efficiency algorithm(s), the system 10 modifies the objective function in the error minimization of the bundle adjustment process to include the uncertainty of key points based on the covariance matrix as an error model for the 2D and re-projected image points. By reducing the 2D key points and including the uncertainty of key points in the error minimization of the bundle adjustment process the system 10, i.e., execution of the animal efficiency algorithm(s) improve both the number of matching points and the error in final dense 3D reconstruction at least tenfold, leading to much more accurate models without major losses in computational time.

[0086] By using the matching of re-projected 3D points from the ToF sensors 30 instead of (or in addition to) 2D key points, the number of matching points will be increased by tenfold. More particularly, by using a combination of 2D and 3D re-projected points from ToF sensors, the animal efficiency algorithm(s) can rely on a much larger number of corresponding 2D points. These are then assigned descriptors and matched as pairs between consecutive and overlapping images. This causes a very small reduction in the number of pairs, which are then further filtered by epipolar constraints and random sample consensus (RANSAC), since many of these points originated from actual 3D points, only a small number are actually eliminated. While mismatches using descriptors may occur, the matching pairs are reconstructed in space and combined with the original 3D points from the 25957884.v1 - 30 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 ToF sensors 30 for a much denser point cloud, which are co-registered through a two-step process involving iterative closest points (ICP) and bundle adjustment.

[0087] In various embodiments, the bundle adjustment step in SfM calculation executed by the animal efficiency algorithm(s) was modified. In practical image measurements, the same 3D feature points may present different distributions of 2D gray-scale patterns depending on the camera position and orientation, affecting the extracted coordinates of key points. Therefore, uncertainty of image point error should be used to represent this anisotropy and non- independent identical distribution of error. In the present disclosure, the inverse of the covariance matrix of measurement error is used to model the uncertainty of the 2D and re-projected 3D key points in the minimization of the error objective function in bundle adjustment. That is, the following equation is implemented:

[0088] u and v directions; andω is the sum of the gray levels of the pixels in the elliptical / circular region א centered at (ui,vi) on the image plane, to compute ^^^^^^= ‖^^. ^^^, ^^^ − ^^. ^^^^, ^^^^‖^, where ^^= Λ ^^ / ^ ^ ^^^^and Λ^^ , ^^^are the eigenvalues and eigenvectors of ^^^, representing the value and orientation of the uncertainty of key point i with pixel coordinate^^^, ^^^and expected coordinate ^^^^, ^^^^ – i.e., from re-projection using as arguments of the minimization the rotation and translation (pose) of the camera. In exemplary test results, the error in spatial coordinates of the final 3D model were reduced by half. Additionally, similar to previous efforts in plant phenotyping, data extracted from the infrared ToF sensor 30 can be co-registered with 3D models to create combined 4D models of cattle. In various embodiments, the system 10 can allow 4D measurements to be correlated across time, creating a 5D model of the scanned subject.

[0089] Experimental Testing: maintenance energy expenditure was determined by using the scanning system together with direct measures of maintenance energy expenditures in 96 cattle. Direct measures of maintenance energy expenditure was accomplished via open-circuit calorimetry, and measures 25957884.v1 - 31 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 of surface area obtained were standardized using computer vision by directly measuring surface area with a surface area integrator.

[0090] Animals were housed in partially covered pens and the pens were equipped with semi-autonomous individual feed monitoring systems. Cattle were utilized to evaluate the relationship of computer-generated surface area to measures of fasting heat production from the standard age of marketing for growth (i.e., weaning) until marketing for beef production. Measures were collected across five stages of growth 612 ± 47, 748 ± 77, 978 ± 111, 1060 ± 109 and 1363 ± 114 lbs. At each growth point, cattle were fasted prior to measures of fasting heat production through indirect calorimetry. Computer-generated surface area of cattle were then determined within 48 hours after measures of fasting heat production.

[0091] Cattle were enrolled into the experiment in groups of 12 when they achieved a body weight of about 600 lbs. After enrolling the first group into the experiment, subsequent groups of calves (initial body weight = 600 lbs) were enrolled every 14 days to facilitate measures of maintenance energy requirements using indirect calorimetry. Measures of surface area using a cattle scanner (such as the system 10) and a surface area integrator were taken on each animal at the beginning of the experiment and when animals weigh 1,000 and 1,300 lbs, respectively. Measures of body weight were measured on the whole animal when body weight estimates from front-end weight were within 50 lbs of the target weight and measures were collected when each cohort has an average weight of 700, 1,000 and 1300 lbs.

[0092] Subsequently, after the cattle were harvested the hide was retained and measures of the surface area were determined by transferring the shape to a common plastic material and calculated after determination of the dry weight of the plastic material.

[0093] Ultimately, these data facilitated direct measures of each method of surface area measures in live cattle. Furthermore, these data allowed development and validation of determination of the relationship between maintenance energy requirements in a range of growing cattle. Estimates of energy retained in the body together with the direct measures of maintenance energy requirements allow for investigation in the relationship between surface area, and 25957884.v1 - 32 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 in various instances volume, maintenance energy requirements and efficiency of energy use to support gain.

[0094] The same cohort of cattle was used to better define the relationship between surface area, maintenance energy expenditure and feed intake among cattle fed forage- or grain-based diets. Feed intake by ruminants is complex, differences in intake are associated with differences in the efficiency with which cattle use feed for productive purposes. Current models of intake in cattle predict that cattle consume feed to a constant energy end-point. However, ruminal fill mechanisms can limit intake in cattle fed forage-based diets. Consequently, current models poorly predict intake of forage fed cattle, and current estimates of intake for cattle are poor at predicting individual differences in intake even when chemostatic mechanisms regulate intake. The primary variables used to predict intake of cattle are metabolic body size and energy density of diets. Thus, it is likely that current inaccuracies in prediction of feed intake are related to the inherent inaccuracies in metabolic body size. Because volume and surface area of objects are directly related, it is envisioned that measures of surface area will allow accurate estimates of feed efficiency and feed intake when cattle consume either grain- or forage-based diets.

[0095] To better define the relationship between surface area, feed efficiency and feed intake in cattle, each pen of cattle were fed either a grain-based diet (e.g., 60% corn grain, 30% finely ground wheat straw, 8% corn gluten meal, 2% vitamins and minerals) or forage-based diet (e.g., 65% corn silage, 33% distillers’ grains with solubles, 2% vitamins and minerals) until a body weight of 800 lbs was achieved. Subsequently, half the pens that received a grain-based diet during the initial phase of the experiment were provided a forage-based diet from 800 lbs of body weight to 1300 lbs of body weight, and the other half of cattle housed in pens receiving a grain-based diet continued to be provided a grain based until the end of the experiment. Alternatively, half of the pens that received a forage-based diet during the initial phase of the experiment were provided a grain-based diet from 800 lbs of body weight to 1300 lbs of body weight, and the other half of cattle housed in pens receiving a forage-based diet in the initial phase of the experiment were provided a forage-based diet until the end of the experiment. Thus, dietary 25957884.v1 - 33 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 treatments were delivered across pens in a 2 × 2 factorial design. Measures of feed efficiency were collected daily from daily measures of individual intake of each animal together with daily estimates of weight gain. Additionally, rates of weight gain were also determined by collecting complete body weights each 28 days together with measures of surface area using the scanner during the experiment.

[0096] Because intake in cattle is regulated by either chemostatic or ruminal fill mechanisms the forage-based and grain-based diets were formulated to contain similar amounts of energy density (e.g., net energy for gain ~ 0.55 Mcal / lb; net energy for maintenance ~ 0.83 Mcal / lb). The observations allowed measures of surface area, feed efficiency and feed intake when intake was controlled by separate biological factors, but were not confounded with differences in energy concentration between diets, which itself could influence differences in measures of feed intake and efficiency. At the conclusion of the experiment, these data were used to enable predictions of intake based on surface area or volume of cattle. Thus, these data provided predictive equations that can accurately estimate the potential for intake in cattle both when intake is limited by chemostatic or physical fill mechanisms.

[0097] To identify genetic parameters, genomic associations and genetic correlations for 4-D image data (including maintenance energy) and relevant production traits, the 96 cattle described above were phenotyped using camera surface area, post-slaughter hide surface area, skin temperature, volume, maintenance energy expenditure directly determined by open-circuit calorimetry, efficiency of energy retained for growth from measures of post-harvest carcass composition, daily feed intake, and daily weight. To maximize the value of these phenotypes, the 96 cattle were genotyped (i.e., DNA tested) using a low-pass sequencing approach. This approach produces highly accurate genotype calls at 70 million DNA variants (most of which are single nucleotide polymorphisms, SNPs). Further, because the entire DNA sample was sequenced, pathogen sequences in the sample were searched for that may be indicative of infection differences between the animals..

[0098] Experiments contrasting high feed intake versus low feed intake using RNA-seq to look at differential expression were completed for various 25957884.v1 - 34 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 beef breeds as well as in other species such as swine. When maintenance energy expenditure is regressed on surface area using the 96 animals at 600 lbs (as described above), the 4 animals with the largest positive residuals were identified, 4 animals with the largest negative residuals were identified, and 4 animals with the smallest residuals were identified. In other words, animals divergent for maintenance energy relative to their surface area (high maintenance, low maintenance, and average maintenance) were identified. Whole blood were sampled for all 96 cattle, but Total RNA Stranded Library (rRNA depletion) libraries were only be prepared for the 12 selected cattle. RNA-seq libraries were prepared and sequenced to produce approximately 50 million reads per sample. Differential expression used community accepted practices, such as a pipeline using fqtrim, Hisat2, FeatureCounts, and edgeRrobust. This experiment identifie genes differentially expressed between animals with high versus low maintenance energy expenditures, illuminating basic biology of maintenance energy. Further, this RNA- seq data was used in a unmapped reads metagenomic approach, as done with the low pass sequencing data.

[0099] In addition to the 96 animals were scanned via a scanning system 10 to obtain 4-D image data, 2,710 other cows were scanned via the system 10 that have been previously genotyped using various known methods. Since the system 10 is portable, as many of these cattle as possible were scanned resulting in 4D image data for at least 4,000 cattle. Animals were imputed to a common set of 835,947 biallelic SNPs. Genome-wide association studies were completed on these 4,000 cattle for surface area, volume, surface area-to-volume ratio, and effective surface area (index of surface area and temperature) using a linear mixed model following community standards for analyses and multiple testing correction. Data was adjusted for age, sex, and contemporary group. Heritability estimates were obtained from these linear mixed models.

[0100] These 4D image data were combined with previously completed datasets to estimate genetic correlations between surface area, volume, and effective surface area (index of surface area and temperature) with various production traits. Genetic correlations were estimated. The 4D image data was fit in bivariate models with feed intake (n = 11,494), average daily gain (n = 11,494), 25957884.v1 - 35 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 carcass weight (n = 16,718), ribeye area (n = 5,224), fat thickness (n = 5,224), marbling (n = 5,224), body weight (n = 11,494), hair shedding (n = 12,000), Bovine Respiratory Disease status (n = 2,074), pelvic area (n = 5,000), and heifer pregnancy (n = 5,000).

[0101] Cattle body weight scales have become an essential part of beef seedstock operations, especially those who register cattle in breeds that require whole-herd reporting of weaning weights. The above testing resulted in strong statistically significant relationships between maintenance energy expenditures, surface area and volume. Accordingly, it is envisioned that the system 10 will become as common place as scales at seedstock farms and ranches. Genetic correlations estimated between various production traits and surface area, volume, and maintenance energy derived from the system 10 can be used to improve the profitability, sustainability, and efficiency of cattle production. More particularly, genomic predictions derived from direct measures of the primary variables (e.g., surface area, volume and maintenance energy) utilizing the system 10 contribute to identifying differences in efficiency in cattle. Furthermore, understanding how 4D image data generated utilizing the system 10 can be correlated with various production phenotypes that will provide for precision cattle breeding. As a result, it is anticipated that genomic prediction models generated utilizing the system 10 will enhance efficiency of cattle and identify animals that produce progeny best suited for optimal economic returns, production efficiency and environmental sustainability across the cattle production system.

[0102] Figure 4 provides an exemplary overview of an autonomous software pipeline implemented by execution of the animal efficiency algorithm(s). After initiation of execution of the animal efficiency algorithm(s), all steps, processes, procedures, calculations, decisions, renderings, outputs, etc., preformed by execution of the animal efficiency algorithm(s) are preformed autonomously. The pipeline begins with the in vivo acquisition of the RGBD data, which undergoes a segmentation and filtering step to eliminate the background pixels and noise in both depth and RGB space. The filtered data is subsequently back-projected into 3D space and then stitched to form a unified 3D model. A mesh is then constructed over the 3D point cloud. Thereafter, traits of interest such as 25957884.v1 - 36 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 surface area and volume are extracted utilizing the 3D point cloud. In various instances, the animal efficiency algorithm(s) have 4 major components; acquisition, processing, registration and trait-extraction.

[0103] Acquisition: Once a live animal is position within the animal retention structure 14, the scanning process starts with the acquisition of the RGB data and the ToF data from all the RGB cameras 26 and all the ToF sensors 30. These data are combined by the animal efficiency algorithm(s) to provide RGBD (color and depth) images (e.g., 3D images) of the animal. The animal efficiency algorithm(s) is / are user-friendly and provide operators the ability to customize settings for the image acquisition process, such as automated or manual assignment of cattle IDs. The algorithm(s) establish(s) connections (wired or wirelessly) with all the data acquisition device control modules 46. In various embodiments, each data acquisition device control module 46 host a server program that processes requests from the user and controls the data acquisition process at a software level. In various embodiments, the animal efficiency algorithm implement ZeroMQ messaging protocol Socket facilitate connections to enable communication between the system controller 42 and the data acquisition device control module server programs. Following image collection, all RGB and depth images are saved locally on the data acquisition device control module 46 along with each data acquisition device’s intrinsic parameters. These images are then transmitted to the system controller 42 for subsequent processing.

[0104] Processing: Next, RGB and depth images are processed to segment the animal’s body, a crucial step for subsequent 3D registration. Initially, we apply a filtering technique, effectively mitigating the "flying pixel" artifacts typical of ToF sensors such as ToF sensors 30. Following this, a Mask R-CNN (Region- based Convolutional Neural Network) is employed, which is a state-of-the-art deep learning model renowned for its efficacy in image segmentation. The unified architecture of Mask R-CNN allows for the accurate identification and localization of multiple objects within complex scenes, making it an ideal solution for tasks demanding a detailed understanding of visual data. However, the software pipeline of the present disclosure modifies and enhances the Mask R-CNN framework, which is traditionally trained soley on RGB images from the COCO (common 25957884.v1 - 37 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 objects in context) dataset. Particularly, dual-modality training by fine-tuning the model on both RGB and depth data is employed. This significantly enhances the model’s capacity to accurately segment cattle within our application-specific dataset. By integrating depth data, the modified and enhanced Mask R-CNN of software pipeline executed by the animal efficiency algorithms of the system 10 not only retains its original segmentation capabilities but also gains an increased sensitivity to the spatial nuances critical to effective 3D scanning in agricultural settings.

[0105] Registration: After segmentation, the 2D coordinates of the segmented object are back-projected into 3D space, resulting in a segmented point cloud for each sensor. These point clouds are then stitched and registered to construct a 3D model (e.g., a 360° surface 3D model) of the cattle as illustrated in Figure 5. Figure 5 is an exemplary multi-view point cloud registration, wherein (a) exemplarily illustrates N = 6 point clouds, and a simple pairwise registration of point cloud fragments of the scanned cattle is performed, and wherein (b) exemplarily illustrates the use a Colored Iterative Closest Point ICP algorithm to solve for the coordinate transformation from camera coordinate frame j to camera coordinate frame i denotediHj). Each view is aligned into the coordinate frame of it’s adjacent camera. The coordinate frame of camera 1 (V1) is fixed as the world coordinate frame and then all views with respect to coordinate frame 1 are aligned. (c) exemplarily illustrates a resulting well aligned point cloud of the scanned cattle. The task of registering all the point clouds is done in a pairwise manner, utilizing the multi-scale the ICP algorithm with an initial alignment. The initial alignment of the point clouds is facilitated through the use of a specially designed calibration cube. This cube is embedded with AprilTag fiduciaries, a type of 2D barcode that provides reliable and accurate 3D pose estimation. This initial alignment sets the stage for the more refined multi-scale colored ICP algorithm. Subsequent to this initial alignment, the point clouds from each pair are registered using the colored ICP algorithm. The ICP algorithm iteratively minimizes the difference between the points of two point clouds, simultaneously taking into account the color and geometric information. Notably, this multi-scale approach progressively refines the alignment starting from a coarse scale and gradually moving to a finer one, thereby ensuring a globally optimal solution while avoiding local minima. 25957884.v1 - 38 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0106] Upon successful pairwise registration, all the registered pairs are linked together. This process effectively consolidates the registered point clouds from each individual RGB camera 26 perspective into one unified, globally consistent coordinate frame. The final, holistic point cloud brings together the different views, providing a comprehensive 3D representation of the object of interest. Figure 5c shows a registered point cloud of an animal after registering all the views. To further refine the 3D model, DBSCAN clustering can be employed, which effectively removes small, potentially erroneous point clusters. The result of this is a well segmented point cloud ready for surface reconstruction. Reconstruction / Trait-extraction: The pipeline concludes with a 3D surface reconstruction of the registered point cloud, facilitated by a Poisson Surface Reconstruction algorithm. The Poisson Surface Reconstruction algorithm is designed to reconstruct a smooth, watertight surface from a disorganized collection of 3D points, enabling the animal efficiency algorithm(s) of the system 10 to extract any traits of interest, for example extract surface area and volume estimates of the animal, as exemplarily illustrated in Figure 3. While the current application primarily focuses on measuring the volume and surface area of cattle, this stage of the pipeline is designed to be adaptable, accommodating the extraction of various phenotypes according to user specifications. This flexibility allows users to define and extract additional traits of interest, extending beyond volume and surface area. Thus, although the utility of the process described above is exemplarily demonstrated to generate specific measurements, the system 10 is broadly applicable and can be tailored to meet the diverse needs of various different users. Additionally, it is envisioned that the scanning methods and procedures described above can be repeated at various different growth stages of each animal (e.g., small, medium and large) track and record maintenance energy requirements of each animal at the various different stages of growth. Table 1 below exemplarily illustrates the software pipeline described above. Table 1 Scanning Software Pipeline 25957884.v1 - 39 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0107] Experimental Results: The below experimental testing and results demonstrate the effectiveness of the scanning system 10 in terms of accuracy and throughput. The experiments were designed to highlight: 1) the impact of effective synchronization on the quality of the scanning results; 2) the precision of scanning system 10 in estimating volume and surface area, by varying object position and orientation, object sizes, lighting conditions, and the distance to the RGB cameras 26 and Tof sensors 30, and : 3) tests on real cattle to evaluate the performance of the system 10 in terms of both accuracy and scanning speed.

[0108] The primary objective of these experiments is to demonstrate the crucial role of ToF sensors (e.g., ToF sensors 30) synchronization in optimizing data quality. The ToF sensors (e.g., ToF sensors 30) operate by projecting modulated near-IR illumination onto a scene, and in a multi-sensor configuration, the potential for signal interference is high if emissions are not carefully timed. In an experimental scenario, ten (10) ToF sensors were arranged pairwise along a 25957884.v1 - 40 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 rod, and two differently sized boxes were used as scanning targets. A synchronization strategy was implemented, introducing a sequential delay of 160µs between each sensor’s activation. Scans were conducted in two distinct operational modes: with and without hardware synchronization, capturing images from ten different perspectives (N=10). The qualitative impact of synchronization is profoundly demonstrated by the data: for instance, as shown in Figure 6, camera 9 captured 37,583 points when synchronized; without synchronization, the point count drastically dropped to 17,098 for the same camera. Notably, as illustrated in Table 2, the master camera (Camera 1) returned a slightly higher point count when unsynchronized. This is expected since Camera 1 always fires first in both synchronized and unsynchronized capture modes, thus avoiding interference from other sensors. These results, illustrated in Table 2 and Figure 6, underscore that improper synchronization can lead to a significant loss of data—more than 50% of the object’s points in some cases. This stark contrast highlights the critical necessity of precise synchronization to prevent substantial data degradation and to ensure the reliability of multi-sensor ToF systems. Figure 6 is an exemplary illustration of comparison of 3D point cloud capture quality with and without synchronization using a large box with known dimensions. The left image displays the results without synchronization (0us), capturing a total of 17,098 points. The right image shows the same box captured with synchronization (160us) with all other settings the same, resulting in a total of 38,631 points, illustrating the significant improvement in data acquisition quality. Table 2 Measurements Of The Number Of Points Generated In The Synchronized And Unsynchronized Mode Of The ToF Sensor 25957884.v1 - 41 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0109] Experiments On Volume And Surface Area Accuracy Using Known Objects: Several experiments were conducted to validate the accuracy of the system 10 in determining the target object’s volume and surface area. Four different objects, including three boxes and one cylinder with known dimensions, were used for testing.

[0110] Single Object, Multiple Orientations: To demonstrate that the behavior of the system 10 is consistent with respect to the object’s pose, a test using a cylindrical object positioned in various configurations was conducted. Both the translation and orientation of the cylinder were varied, placing it in eight (8) distinct positions. Images were captured from ten (10) different RGB cameras (e.g., RGB cameras 26) for each position. The acquired RGB-D images were subsequently stitched together and then a mesh model of the cylinder was generated. The surface area and volume were calculated based on this mesh. The results, presented in Figure 7, confirm that the system 10 consistently and accurately computed the surface area and volume of the cylinder across all tested orientations. Specifically, the mean surface area was 0.725 m2(averaged over eight poses), closely approaching the ground truth of 0.73 m2with a narrow standard deviation of 0.009 m2. For volume, the mean value was 0.027 m3(averaged over eight poses), nearly matching the ground truth of 0.03 m3with an exceptionally low standard deviation of 0.0008 m3. These results not only affirm the precision of the system 10 measurements but also validate the robustness of the system 10 across diverse orientations. Figure 7 exemplarily illustrates results of scanning a cylindrical object in multiple orientations, highlighting the accuracy across diverse poses of the 25957884.v1 - 42 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 system 10. The horizontal axis displays the predicted volumes and surface areas obtained in each test. Given that the same object was used throughout, the ground truth volume and surface area remain constant. This plot demonstrates the precision of the system 10, as evidenced by the close alignment of the predicted values with the consistent ground truths, illustrating the reliability of the system 10 in varying orientations.

[0111] Multiple Objects, Static Pose: To further evaluate the performance of the system 10, a series of tests focusing on its ability to consistently predict surface area and volume across a range of objects differing in size and shape was conducted. The test suite included a cylinder, a small box, a medium box, and a large box, each with precisely known dimensions. Given the demonstrated pose-independence of the system 10, all objects were placed in the same central position and subjected to 10 consecutive scans without altering their pose. The average predicted surface area and volume from these scans were computed and a regression analysis to assess the precision of the measurements was performed. The fit through the predicted volume and surface area data yielded an R2value of 0.997 for surface area and 0.999 for volume, indicating extremely high accuracy and consistency. The results, displayed in Figure 8, underscore the robustness and reliability of thesystem 10. The system 10 not only predicts volume and surface area accurately across diverse object types but also demonstrates excellent repeatability. These findings confirm the efficacy of the system 10 in various operational scenarios, enhancing its applicability in real-world conditions. Figure 8 exemplarily illustrates regression analysis of predicted versus known surface area and volume for multiple static objects. The plots display the correlation between values predicted via the portable animal retention, data collection and analysis system shown in Figure 1 and the actual measurements for a cylinder, small box, medium box, and large box, all placed in the same pose across 10 consecutive scans. The high R2values of 0.997 for surface area and 0.999 for volume demonstrate the scanner’s accuracy and consistency in various object dimensions and shapes under controlled conditions.

[0112] Scanner Performance under Sunlight: To assess the robustness of the system 10 in realistic operational environments, particularly for livestock (cattle) 3D scanning, an experiment in an outdoor setting under direct 25957884.v1 - 43 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 sunlight was conducted. This test aimed to determine the impact of sunlight on the accuracy of surface area and volume measurements. A single box was used for this purpose, with 10 consecutive scans performed, via the system 10, to ensure consistent results. The processed data from these scans were stitched together to create a mesh from which the surface area and volume were computed. The results indicate a mean volume of 0.10911 m3, with the ground truth being 0.125 m3, a 12.71% difference, with a standard deviation of 0.0090 m3. For surface area, the mean measurement across the scans was 1.393 m2, compared to the ground truth of 1.500 m2, a 7.13% difference with a standard deviation of 0.0999 m2. These findings, depicted in Figure 9, highlight the ability of the system 10 to perform under challenging lighting conditions, though they also suggest a slight decrease in measurement accuracy due to the effects of sunlight on the infrared ToF sensors 30. This insight is crucial for optimizing scanner deployment in outdoor livestock scanning scenarios. Figure 9 exemplarily illustrates the performance of the system 10 under direct sunlight, using a standard box to simulate outdoor livestock scanning conditions. The graphs show the mean and standard deviation of volume and surface area measurements across 10 consecutive scans. The results here illustrate the slight impact of sunlight on the infrared Tof sensors 30 of the system 10, affecting measurement accuracy.

[0113] Varying Object Distance to the Sensor: Further experimentation, explored how the distance between the RGB cameras / ToF sensors and an object influences the measured surface area and the number of points captured by the sensor. An object was positioned at incremental distances, starting from 2 feet up to 7 feet, from an RGB camera and ToF sensor (e.g., an RGB camera 26 and ToF sensor 30) ToF measurements were taken at each foot interval using both a narrow field of view (NFOV) and a wide field of view (WFOV) settings of the ToF sensor. The results, as depicted in Table 3, indicate that the surface area measurements and the number of points captured varied with distance, but not in a linear fashion. For both NFOV and WFOV, the maximum surface area was recorded at a distance of 3 feet with values of 0.349 m2and 0.343 m2respectively. Notably, both the surface area and the number of points tended to decrease as the object is positioned further from 3 feet, reaching their lowest at 7 feet with surface area 25957884.v1 - 44 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 values of 0.251 m2for NFOV and 0.254 m2for WFOV. Interestingly, the measurements do not show a consistent decrease as distance increases, which suggests that the optimal distance for scanning in terms of maximizing the captured surface area and point density using a ToF sensor 30 is around 3 feet. Beyond this point, both metrics begin to decline, possibly due to the spreading and weakening of the sensor’s signal with increased distance. These findings highlight the importance of the placement of the ToF sensors 30 relative to the target animal in capturing the most accurate and detailed data possible, via the system 10. Table 3 Statistical Results Of The Varying Of The Distance Of An Object From the ToF Sensor

[0114] Experiments on Real Cattle: Following the validation of the scanner on objects with known dimensions, experiments on real cattle were conducted. The initial step in the experimental pipeline involved fine-tuning the segmentation network to adapt it specifically to cattle data. For this process, testing captured both depth and RGB images using RGB cameras (e.g., RGB cameras 26). Two pre-trained Mask R-CNN models were fine-tuned using these image types respectively. To determine the most effective segmentation approach, a voting arbitration mechanism between the two models was employed. The outcomes of four different voting arbitration strategies are presented in Table 4, which exemplarily illustrates the statistical results (the average IOU, the false positive rate (FR), and the false negative rate (FN)) of using different voting arbitration for the segmentation models. Table 4 25957884.v1 - 45 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026Intersection over Union (IOU) score across all arbitrations, suggesting strong segmentation performance. However, this model also exhibited a high rate of false negatives, which could result in the loss of essential surface details crucial for accurate surface area computation of the cattle.

[0116] In contrast, the single-vote arbitration method, combining votes from both the depth and RGB models, provided an IOU score comparable to the RGB-only model, but significantly reduced the false negative rate. Consequently, this method was selected for further segmentation tasks. Qualitative results of the segmentation on the test set are exemplarily illustrated in Figure 9. Figure 9 exemplarily illustrates the segmentation of cattle using combined RGB and depth models via Mask R-CNN during execution of the software pipeline of the animal efficiency algorithms: Figure 9 shows an RGBD image of cattle segmented using both RGB and depth data. Results from each model are integrated using a voting arbitrator, resulting in a well-defined segmentation in both modalities. After successfully detecting and masking the animals within the RGB and depth images, the generated masks were used to precisely extract the animals from the overall image. This segmentation isolates the animal alone, facilitating further processing of the data.

[0117] Next, surface and volume measurements were extracted for the cattle. Table 5 shows the statistics of the experiment conducted on live cattle versus their hides post-slaughter. Hides were removed at variable distances from the head and hooves. More particularly, Table 5 exemplarily illustrates a comparison between the measured surface area using the system 10 described herein and the hand measurements from the hide of slaughtered animals. Each of the animals (which is determined by its cattle ID) was scanned 5 times via the system 10 and the average value was calculated over the scans. 25957884.v1 - 46 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 Table 5

[0118] Thebased scanner system optimized for high-throughput cattle phenotyping. The system 10 proficiently captures and reconstructs meshes from various objects, demonstrating robustness across multiple challenging conditions. Through testing such as that described above, it has been shown that the system 10 maintains consistent accuracy irrespective of the object’s geometry, orientation, position, distance to sensor and exposure to environmental elements such as sunlight. In real world applications, direct sunlight can be minimized using tents or other apparatus. The system 10 demonstrated remarkable resilience in handling both stationary and untamed cattle, a capability enhanced by the animal efficiency algorithms working in tandem with hardware synchronization. This synchronization effectively mitigates the issue of time-of-flight (ToF) sensor interference, which is crucial for accurate data acquisition. The design of system 10 includes a heavy- duty cattle chute mechanism as described above, which, together with timely camera triggers, guarantees reliable data capture from dynamic subjects. The entire process, from capture to the computation of mesh measurements like volume and surface area, is completed within an average time of approximately 10.96 seconds. This efficiency underscores the system’s suitability for high-throughput applications, making it a valuable tool for modern agricultural practices. 25957884.v1 - 47 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0119] Referring now to Figures 1 and 11, as described above, in various embodiments, system 10 can be a wireless system wherein the data acquisition system 18 communicates wirelessly with the data processing system 22. Figure 11 illustrates an exemplary embodiment of such a wireless system 10. In such embodiments, the data acquisition system 18 comprises the plurality of data acquisition devices 24 that are positioned near or adjacent to the retention structure 14 such that their combined fields of view or lines of sight create a volume large enough to cover the entire animal positioned within the retention structure 14, as described above. Additionally, in such embodiments the data acquisition devices 24 can comprise a plurality of RGB cameras 26 and a plurality of infrared time-of- flight (ToF) cameras / sensors 30, and in some instances a plurality of thermal sensors / cameras 34, as described above. Moreover, in such embodiments, data processing system 22 comprises the system controller 42 and data acquisition device control modules 46, as described above, only each data acquisition device control module 46 is located near or in close proximity to the respective data acquisition devices 24. It is envisioned that in various implementations the data acquisition device control modules 46 can be mounted to or are integrally formed with one or more of the data acquisition devices 24. The system controller 42 is structured and operable to wirelessly communicate with each of the data acquisition device control modules 46.

[0120] Importantly, in such wireless embodiments, the system 10 is structured and operable to implement a wireless synchronization of the RGB cameras 26 and the ToF sensors 30. More specifically, with the wired embodiment of the system 10, the ToF sensors 30 require a signal to synchronize with the respective RGB camera 26 and thereby synchronize their image acquisitions. Failure to do so can cause great loss in the data acquisition, which affects the 3D model reconstructed, and hence the quality of the volume and surface area estimations generated as described above. In various wireless embodiments of the system, such as the system 10 exemplarily illustrated in Figure 11, wireless synchronization has been implemented that eliminates the need for running synchronizing wires between the RGB cameras 26 and the corresponding ToF sensors 30. Specifically, in such wireless embodiments, the system 10 the data 25957884.v1 - 48 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 acquisition devices 24 further include a plurality of Internet-of-things (IoT) devices 50. Each IoT device 50 is mounted to or located near or in close proximity to a respective ToF sensor 30 and RGB camera 26. The IoT devices 50 provide both WiFi and I / O capabilities such that each IoT device 50 can be wirelessly interconnected using the same WiFi hub already available for data acquisition control modules 46 of the system. Therefore, each IoT device 50 can wirelessly communicate with both the respective corresponding data acquisition control modules 46 and the system controller 42. Accordingly, each IoT device 50 is structured and operable to control the synchronization of the operation of the RGB cameras 26 and the respective corresponding ToF sensors 30. Furthermore, a program with a broadcasting protocol is implemented into each IoT device 50, and the software pipeline described above that is executed by the system controller 42 is configured to trigger a unique broadcast message that causes each and every IoT device 50 to coordinate image capture in a sequence (i.e. at different times), guaranteeing no major interference between the plurality of ToF sensors 30. Particularly, the synchronization is implemented as follows: 1) first, each IoT device 50 is mounted to or located in located close proximity to a respective ToF sensor 30 and RGB camera as described above; 2) next, a short wire is also connected from a I / O port of the IoT device 50 to the synchronization input of the respective ToF sensor 30; 3) next, at the beginning of each scanning process (i.e., RGB image data capture), the system controller 42 sends a broadcast message to all IoT devices 50, and each IoT device 50 pulses their synchronization signal outputs (I / O), which causes the ToF sensors 30 to be in perfect synchronization.

[0121] Referring now to Figures 12, 13, 14 and 15, as described above, the system 10 comprises the animal retention structure 14 that is structured and operable to temporarily restrain an animal (e.g., a cow) in specific location so that the data acquisition system 18 and data processing system 22 can acquire data of the animal (e.g., surface area, surface temperature and / or volume) and then measure individual differences in efficiency in the various animals, as described herein. In various embodiments, the animal retention structure 14 comprises a main chassis or frame 54, an animal head yoke 58 and a pair of opposing side fences 62 (e.g., a first side fence 62 and a second side fence 62). The main chassis 54 is 25957884.v1 - 49 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 constructed to have a height, a width and a length and is sized to receive and enclose or envelope an animal to be scanned utilizing the system 10. The head yoke 58 comprises a first yoke gate 58A and a second yoke gate 58B that are movably (e.g., hingedly or slidingly) connected to a front end 54A of the chassis 54 such that the head yoke 58 is selectively moveable between an Open position (as shown in Figure 13) and a Closed position (as shown in Figures 12, 14 and 15). A rear end 54B of the chassis 54 is structured to attach to, or otherwise be disposed adjacent, an animal chute (not shown) that guide the animals into the animal retention structure 14 for scanning as described herein. Each side fence 62 comprises a first side fence gate 62A and a second side fence gate 62B that are movably (e.g., hingedly or slidingly) connected to a respecting one of opposing sidewalls 54C of the chassis 54 such that each side fences is selectively movable between an Open position (as shown in Figures 14 and 15) and a Closed position (as shown in Figures 12 and 13).

[0122] Generally, in operation, initially the animal retention structure is configured to have the head yoke 58 and both side fences 62 are all in the Closed position (as shown in Figure 12). Subsequently, animals enter the animal retention structure 14 through the rear end 54B of the main chassis 54, which can be attached to, or otherwise disposed adjacent, an animal chute. Hence, an animal can be herded directed or guided from a corral, a truck or trailer through an animal chute and into the animal retention structure 14 for scanning. As the animal moves forward into the animal retention structure 14 the first and second yoke gates 58A and 58B are moved to place the head yoke 58 in the Open position (Figure 13). With the head yoke 58 in the Open position, the animal can extend its head past the front end of the chassis 54 and between the first and second yoke gates 58A and 58B. The first and second yoke gates 58A and 58B are then moved to place the head yoke 58 in the Closed position (Figures 12, 14 and 15) whereby the first and second yoke gates 58A and 58B are disposed adjacent and in close proximity to the animal’s neck such that the first and second yoke gates 58A and 58B restrain the head of the animal and the animal cannot pull its head back out between the first and second yoke gates 58A and 58B. Accordingly, the animal is temporarily restrained in a scanning position within the animal retention structure 14. 25957884.v1 - 50 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0123] Once the animal is restrained in the scanning position within the animal retention structure 14, the first and second fence gates 62A and 62B are moved from the Closed position to the Open position. In the Closed position each of the first and second fence gates 62A and 62B are disposed against and / or latched or otherwise secured to the at least a portion of the chassis sidewalls 54C, and at least partially obstruct the fields of view and / or the lines of sight of the data acquisition devices 24. However, once the animal is restrained in the scanning position, the first and second fence gates 62A and 62B are moved out fields of view and / or the lines of sight of the data acquisition devices 24 to place the opposing side fences 62 in the Open position (Figures 14 and 15). Moving the side fences 62 to the Open position provides an unobstructed, nonoccluded, or un-interfered fields of view or lines of sight for any and all data acquisition devices 24 positioned to capture data of or for the sides of the animal, as shown in Figure 15. Consequently, the data acquisition system 18 and data processing system 22 are operated as described above. With the side fences 62 in the Open position accurate and clear image and other data can be captured by the data acquisition system 18 and processed by the data processing system 24. Once the image and other data are captured, the first and second fence gates 62A and 62B are moved to place the opposing side fences 62 in the Closed position (Figures 12 and 13) and the first and second yoke gates 58A and 58B are moved to place the head yoke 58 in the Open position (Figure 13) to allow the animal to be removed from the animal retention structure 14. Thereafter, the process can be repeated on the next animal. With the head yoke in the Open position.

[0124] Referring now to Figure 16, it is envisioned that in various embodiments, the animal retention structure 14 can be absent the opposing side fences 62. In such embodiments, it is further envisioned that the opposing sidewalls 54C of the chassis 54 can be constructed to provide one or more large window or opening 66 that provide unobstructed, nonoccluded, or un-interfered fields of view or lines of sight for any and all data acquisition devices 24 positioned to capture data of or for the sides of the animal.

[0125] Furthermore, it is envisioned that in various embodiments, the animal retention structure 14 can comprises the chassis 54, or a framework 25957884.v1 - 51 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 structure similar to the chassis 54 that is structured and operable to temporarily restrain an animal (e.g., cattle) therewithin, wherein the chassis 54 or framework structure can be position at, on, adjacent, abutting or near the edge of a body of (e.g., a water trough, a pond, a creek, etc.). In such embodiments, the chassis 54 or framework structure is positioned such that an animal can enter a rear end of the chassis 54 or framework structure (e.g., rear end 54B) that is located away from the body of water in order to drink from the body of water. In such embodiments the data acquisition devices 24 can be disposed, deployed and located around the chassis 54 or framework structure as described and illustrated above. Additionally, in such embodiments, the data acquisition devices 24 can include a pressure sensing pad, plate or platform, a scale, or any other suitable weighing device or system that is positioned within the chassis 54 or framework structure and is located near a front end of the chassis 54 or framework structure (e.g., front end 54A) that is located at, on, adjacent, abutting or near the edge of a body.

[0126] In operation, the animal enters the rear end of the chassis 54 or framework structure and proceeds into the chassis 54 or framework structure to drink from the body of water. While drinking the animal stands with its front feet on the pressure sensing pad, plate or platform, a scale, or any other suitable weighing device or system. During this time an operator initiates the image data and volume data acquisition process described above which also measures how much the front of the animal weighs via the pressure sensing pad, plate or platform, a scale, or any other suitable weighing device or system. Alternatively, as described above, in various instances the animal can be tagged with an electronic ID tag such that when the animal pauses to drink from the body of water, an EID or RFID reader, reads the animal’s electronic ID and initiates the image data and volume data acquisition process described above which also measures how much the front of the animal weighs via the pressure sensing pad, plate or platform, a scale, or any other suitable weighing device or system. Thereafter, execution of the animal efficiency algorithm(s) utilize the image data, temperature data, volume data and / or animal weight data to generate the 3D (or 4D) images to determine the surface area, and in various embodiments the volume and surface area:volume ratio, of the animal, and predict maintenance energy requirements as described above. 25957884.v1 - 52 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026

[0127] It is further envisioned that in various embodiments, the system 10 as described above with regard to Figures 1-16, can also include a pressure sensing pad, plate or platform, a scale, or any other suitable weighing device or system to acquire animal weight data in addition to the image data, temperature data and / or volume data described above. In such embodiments execution of the animal efficiency algorithm(s) utilize the image data, temperature data volume data and / or animal weight data to generate the 3D (or 4D) images to determine the surface area, and in various embodiments the volume and surface area:volume ratio, of the animal, and predict maintenance energy requirements as described above.

[0128] The description herein is merely exemplary in nature and, thus, variations that do not depart from the gist of that which is described are intended to be within the scope of the teachings. Moreover, although the foregoing descriptions and the associated drawings describe example embodiments in the context of certain example combinations of elements and / or functions, it should be appreciated that different combinations of elements and / or functions can be provided by alternative embodiments without departing from the scope of the disclosure. Such variations and alternative combinations of elements and / or functions are not to be regarded as a departure from the spirit and scope of the teachings. 25957884.v1 - 53 -

Claims

Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 CLAIMS What is claimed is: Claims 1. An in vivo animal retention, data collection and analysis system, said system comprising an animal retention structure structured and operable to temporarily restrain a live animal to be scanned in specific location; an animal data acquisition system structured and operable to acquire phenotypic data of the animal to be scanned; and a data processing system that is structured and operable to execute one or more animal efficiency algorithm that utilizes the acquired phenotypic data to calculate the efficiency of the animal to be scanned.

2. The system of Claim 1, wherein the data acquisition system comprises one or more data acquisition device communicatively connected to the data processing system.

3. The system of Claim 2, wherein the one or more data acquisition device comprises one or more red-green-blue (RGB) cameras and one or more time-of-flight sensor.

4. The system of Claim 3, wherein the one or more data acquisition device further comprises one or more Internet-of-things device.

5. The system of Claim 4, wherein the one or more data acquisition device further comprises one or more thermal sensor.

6. The system of Claim 2, wherein the animal retention structure comprises: a main chassis sized and structured to receive and enclose the animal to be scanned; and 25957884.v1 - 54 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 an animal head yoke movably connected to a front end of the chassis, the animal head yoke comprising a first yoke gate and a second yoke gate, wherein the head yoke is movable between an Open position wherein a head of the animal to be scanned can be placed between the first yoke gate and the second yoke gate and a Closed position wherein the head of the animal to be scanned is restrained by the first yoke gate and the second yoke gate 7. The system of Claim 6; wherein the animal retention structure further comprises a pair of side fences movably connected to opposing sidewalls of the chassis, each side fence comprising a first side fence gate and a second side fence gate, wherein each side fence is movable between a Closed position wherein the first and second fence gates at least partially obscure a field of view of the side of the animal for the one or more data acquisition device and an Open position wherein the first and second fence gates are positioned out the field of view of the one or more data acquisition device.

8. The system of Claim 6, wherein the main chassis comprises opposing sidewalls, each sidewall comprising one or more window that provides an unobscured field of view of side of the animal. 25957884.v1 - 55 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 9. An animal retention structure, said animal retention structure comprising: a main chassis sized and structured to receive and enclose a live animal to be scanned; and an animal head yoke movably connected to a front end of the chassis, the animal head yoke comprising a first yoke gate and a second yoke gate. 10 The animal retention structure of Claim 9, wherein the first yoke gate and the second yoke gate are movably mounted to the front end of the chassis such that the animal head yoke is movable between an Open position wherein a head of the animal to be scanned can be placed between the first yoke gate and the second yoke gate and a Closed position wherein the head of the animal to be scanned is restrained by the first yoke gate and the second yoke gate.

11. The animal retention structure of Claim 10 further comprising a pair of side fences movably connected to opposing sidewalls of the chassis, each side fence comprising a first side fence gate and a second side fence gate.

12. The animal retention structure of Claim 11, wherein each side fence is movable between a Closed position wherein the first and second side fence gates at least partially obscure a field of view of one or more data acquisition device and an Open position wherein the first and second side fence gates are positioned out of a field of the view of the one or more data acquisition device.

13. The animal retention structure of Claim 10, wherein the main chassis comprises opposing sidewalls, each sidewall comprising one or more window that provides an unobscured field of view of side of the animal. 25957884.v1 - 56 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 14. A method for in vivo measuring the maintenance energy requirements of animal, said method comprising: autonomously measuring a surface area of a live animal utilizing computer imaging data; and autonomously performing real-time assessment of maintenance energy expenditure of the animal utilizing the computer imaging data.

15. The method of Claim 14, wherein autonomously measuring the surface area of a live animal comprises: acquiring red, green, blue (RGB) image data and depth data from each of a plurality of RGB cameras disposed at plurality of different angles and perspectives of around the animal; filtering the RGB and depth data of each RGB camera to eliminate background pixels and noise; processing the filtered RGB and depth data of each RGB camera to segment the animal’s body; back-project the filtered and segmented data of each RGB camera into three-dimensional (3D) space resulting in a segmented point cloud for each RGB camera data; stitching the back-projected data for each RGB camera together to form a 3D point cloud of the animal; reconstruct a smooth 360° surface of the animal utilizing the 3D point cloud; and determining the surface area of the animal utilizing the reconstructed surface.

16. The method of Claim 15 wherein autonomously performing real-time assessment of maintenance energy expenditure of the animal comprises 25957884.v1 - 57 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 determining the maintenance energy of the animal utilizing the determined surface area of the animal.

17. The method of Claim 16, wherein autonomously measuring the surface area of a live animal further comprises guiding the animal into an animal retention structure of an animal retention, data collection and analysis system to temporarily restrain an animal to be scanned in specific location.

18. The method of Claim 17, wherein autonomously measuring the surface area of a live animal further comprises: moving a head yoke of the animal retention structure to an Open position such that a head of the animal can be placed between first yoke gate and a second yoke gate of the head yoke; positioning the head of the animal between the first yoke gate and a second yoke gate; and moving the head yoke to a Closed position such that the head of the animal is restrained by the first yoke gate and the second yoke gate, thereby restraining the animal in a scanning position withing the animal retention structure.

19. The method of Claim 18, wherein autonomously measuring the surface area of a live animal further comprises moving a first side fence gate and a second side fence gate of a first side fence of the animal retention structure, and moving a first side fence gate and a second side fence gate of a second side fence of the animal retention structure such that the first and second side fences are placed in an Open position wherein the first and second side fences are out of a field of a view of one or more data acquisition device of the animal retention, data collection and analysis system.

20. The method of Claim 19, wherein autonomously measuring a surface area of a live animal utilizing computer imaging data further comprises acquiring the RGB image data and depth data via the plurality of RGB cameras after the first and second side gates have been moved to the Open position.

21. The method of Claim 14 further comprising: 25957884.v1 - 58 -Atty. Dkt. No. UMCO H064WO / 17193-00116 Client Ref. No.22UMC026 autonomously measuring a volume of a live animal; and autonomously performing real-time assessment of maintenance energy expenditure of the animal utilizing the computer imaging data and the volume data. 25957884.v1 - 59 -

Images