NCSU Extension Swine Husbandry 2004
A more printable version of Swine News in Adobe Acrobat.


May, 2004 . Volume 27, Number 04

The Science of Odor

PRODUCTION CONTROL SYSTEMS FOR PIGS

Growing pigs for meat is little different from other production processes:

  • The rate of production (daily live weight gain) relates to the rate of input of the primary raw material (daily feed supply).

  • The efficiency of production of the output is a function of the total quantity of input resource used (kg feed used per kg pig sold).

  • Inefficiencies result in the production of waste (such as ammonia), which leads to disposal costs.

  • The process is limited by the availability of the first limiting resource, which will likely be one of

    • The quality of feed and of management inputs.
    • The ability of the pig to make use of the inputs presented to it.
  • Management depends upon process control of
    • Production targets.
    • Measurement of the status of the product at any given moment.
    • The possibility of predictable correction of deviation from required performance.

Among management's tools are:

  • Genetic selection of the pig to grow high-quality meat at the required rate and of the required ratio of lean to fat.

  • The daily amount of feed supplied.
  • The balance of energy to amino acid in the diet.
  • The choice of the moment of slaughter (target size).

Effective use of these tools, and the decisions pertaining to their use, is necessarily an on-going, day-by-day process, not least because:

  • The amount of feed the pig requires changes daily as it grows.
  • The balance of energy to amino acids required in the diet changes daily as the pig grows.

  • Both of the above are affected by the ambient environment and the disease environment, which fluctuate on a short time scale.

  • The inherent potential of an individual batch of pigs is not well characterized when the pigs enter the growout unit.

  • All the interacting elements of the production process are subject to short-term biological variation.

Presently, pig production is usually managed retrospectively, not in real time. At worst, pigs are fed according to some general rules (nutrient requirements) pertaining to general populations and measured with pigs from various points in historical time. At best, pigs are fed according to the lessons learned from previous batches reared in like circumstances. Pig production systems, therefore, rarely use the available management tools to control the process and maximize the efficiency of production of meat of the required quality. There are three important reasons for this (Whittemore et al., 2001):

  • The relationships between change in input (amount and quality of feed) and change in output (amount and quality of pig carcasses) are inadequately known at the level of the individual production plant (farm). Such descriptors require modification to fit the conditions prevailing to any given time, place, and circumstance of production. We need accurate, flexible models of pig nutrition and growth.

  • The progress of the process is usually only poorly known. Short-term (and long-term) changes in growth rate are difficult to measure. The conventional (platform) weight scale is not an effective tool for monitoring growth in the short term and is not much used by managers for this purpose. We vitally need an effective way of monitoring and measuring the growth of pigs in real time.
  • The most effective means of controlling the rate of pig growth and the fatness of the end product is by altering the amount and composition of the feed supplied. As the deviations occur in the short term, so also must corrective actions. Feeding systems that allow daily control of both amount and composition of diets fed are presently available. But they are rarely used to their fullest because of inadequate monitoring of present positions of the pigs and insufficient accuracy in predicting the consequences of changes to the pigs' diets. Systems that can deliver variable quantities and qualities of feed on a daily basis are usually programmed to a preset pattern over the whole of the growout cycle. Computer-controlled feeding systems await management tools for their effective deployment.
Visual Image Analysis (VIA)

The visual imaging system uses simple video camera technology to capture the image of the back of the pig as the camera looks down upon the pig. From the image, linear measurements can be taken automatically. Further, and most importantly, mathematical algorithms developed at Silsoe Research Institute (Marchant et al., 1999) allow the outline of the body to be traced and quantified. Thus, the area described by the outline of the pig (A4) can be determined. Early results showed that the relationship between the A4 area and the weigh platform weight was rather close (Figure 1). The original equation for Large White x Landrace pigs was:

Large White x Landrace Live weight = 0.052A4 - 32

Figure 1. Relationship between time (day) and the weight of a pig determined manually and by visual imaging (VIA).

More recently (White et al., 2003), the determination of this relationship was repeated in circumstances differing in time, place, research team and pig type. Three breed types were used: 25% Meishan, 50% Pietrain, and 50% Landrace. The equations describing the relationship between the A4 area as measured by visual imaging and the live weight as measured by a weigh platform were:

50% Landrace Live weight = 0.049(0.0005)A4 - 29.8(0.74)
50% Pietrain Live weight = 0.052(0.0004)A4 - 38.5(0.67)
25% Meishan Live weight = 0.052(0.0004)A4 - 32.8(0.63)

We have interpreted these results as follows:

  • The equations are robust.
  • VIA A4 bears a close relationship to live weight, and the equation can be used to determine live weight.
  • As the equations differ among the pig types, it is important to be aware of pig type when setting up a system.
  • As the equations differ among the pig types, it would appear the VIA system might be able to identify type differences. These could be important for carcass quality issues, such as ham shape and joint proportions.

The most recent work with Large White x Landrace pig types (again completed at a different place and time and with a different team) produced the following:

Landrace x Large white Live weight = 0.049(0.0006)A4 - 34.1(1.15)

which is similar to those determined for Landrace and Large White x Landrace pigs shown above. As all of these relationships were for pigs between 30 and 110 kg live weight, we further tested the relationship with heavier (Large White x Landrace) pigs to find:

80 - 150 kg pigs Live weight = 0.041(0.002)A4 - 33.8(6.0)

It has yet to be ascertained whether the relative loss of precision and determination of slightly differing coefficients may be due to inherent (but stable) differences at higher live weights or perhaps a tendency for curvilinearity in the line.

We believe that determining the status of pigs (size, shape, live weight) by VIA has advantages over a conventional platform weigh scale. Some of these are given in Table 1:

Table 1. Advantages of VIA in comparison to conventional weigh scales

For effective control of the production system, one should know how many days must elapse between two estimates of weight before a safe determination of weight change can be made. Results are given in Table 2, from which it would appear that the VIA system requires only one day more than the weigh scale.

Table 2. Number of days elapsed until expected change in pig size becomes significant, using conventional platform weigh scale and VIA A4 data, assuming confidence limits of 80% and 95%

The (albeit small) between-type differences in the relationship between weight and VIA A4 (below) left us with the possibility that the visual image could be used to sort animals on the basis of their shape. On the assumption that type (25% Meishan, 50% Landrace, 50% Pietrain) was a reasonable description of shape (which would not always be the case due to natural variation within types), a neural network was used to test this hypothesis. Half the animals were presented to the network with a statement of their type together with the full data set for each pig from VIA. This was then used to sort the remaining half on the basis of the VIA data only. The results are presented in Table 3. The accuracy of prediction was high (73%, 83%, and 64%), particularly for the 50% Pietrain and 25% Meishan types. These types were also rarely confused.

Table 3. Confusion matrix for pig type

In conclusion, effective control of the pig production process appears to be hampered by a lack of means to determine accurately, continuously, and in real time a change in pig performance and deviation from targets. The evidence presented suggests the Visual Image Analysis may provide the solution. Management assumes a realistic expectation of outcomes of corrective interference in particular farm circumstances. This can now be provided through medium of a model that can be flexed through (only) two parameters. It would also appear that VIA could inform about pig shape, which may have benefits for carcass appraisal.

The next steps will be to link the model and the VIA system to automatic computer-operated feeding systems, which can control feed level and feed quality. When this step is complete, managers can not only be informed about and diagnose unwanted deviations in the production processes, but like managers of other production processes, they can also take timely remedial action.

—Adapted by Eric van Heugten, from Colin Whittemore, University of Edinburgh, as presented at the 4th London Swine Conference, Building Blocks for the Future, April 1-2, 2004, London, Ontario, Canada.

ON-FARM PERFORMANCE TESTING

The following breeders with validated herds have tested animals in the past 30 days.

Breeder Address Breeds
Bob Ivey* 314 N.C. 111 S, Goldsboro 27530 L, D, H, Y, CW, X
Wesley Looper* 4695 Petra Mill Rd., Granite Falls 28630 L, D, H, Y, X
Thad Sharp, Jr., & Sons 5171 N.C. 581 Hwy., Sims 2788030 D, Y, X
Tidewater Research Farm* 207 Research Station Road, Plymouth 27962 D, Y, L
Thomas Farms 8251 Oxford Rd., Timberlake 27583 X
*Real-Time Ultrasound

—Frank Hollowell and David Lee

Number of accesses to this file:
Last modified April 27, 2004.