Continuous wave metal detector

1. A method comprising:

accessing a training set including a target object set and a clutter object set, the target object set including a target associated with a target feature value and the clutter object set including a non-target object associated with a clutter feature value;

determining that the training set includes multiple types of targets or multiple types of clutter;

comparing the target feature value of a type of target with the clutter feature value;

associating the type of target with the non-target object based on the comparison;

training a classifier using the target feature value and the clutter feature value of the associated type of target and the non-clutter object such that the classifier produces a metric that indicates that an object associated with the type of target is a target;

generating multiple classifiers, the multiple classifiers including the trained classifier;

applying a feature value associated with an unknown object to the multiple classifiers to generate a set of metrics for the unknown object;

aggregating the metrics into an overall metric; and

determining whether the unknown object is included in the target set based on the overall metric.

2. The method of claim 1 , wherein aggregating the metrics into an overall metric comprises summing the metrics included in the set of metrics.

3. The method of claim 2 , wherein, prior to summing the metrics, the metrics are normalized.

4. The method of claim 1 , wherein the multiple classifiers include at least two different types of classifiers.

5. The method of claim 1 , wherein the clutter set includes data representing multiple different types of soils.

6. The method of claim 1 , wherein the target set includes data representing multiple different types of land mines.

7. The method of claim 1 , wherein comparing the target feature value of a type of target with the clutter feature value comprises determining a measure of similarity between the target feature value and the clutter feature value, and associating the type of target with the non-target object based on the comparison comprises associating the type of target and the non-target object when the measure of similarity is below a threshold value.

8. A computer-readable non-transitory medium encoded with a computer program comprising instructions that, when executed, operate to cause a computer to perform operations comprising: accessing a training set including a target object set and a clutter object set, the target object set including a target associated with a target feature value and the clutter object set including a non-target object associated with a clutter feature value;

determining that the training set includes multiple types of targets or multiple types of clutter;

comparing the target feature value of a type of target with the clutter feature value;

associating the type of target with the non-target object based on the comparison;

training a classifier using the target feature value and the clutter feature value of the associated type of target and the non-clutter object such that the classifier produces a metric that indicates that an object associated with the type of target is a target;

generating multiple classifiers, the multiple classifiers including the trained classifier;

applying a feature value associated with an unknown object to the multiple classifiers to generate a set of metrics for the unknown object;

aggregating the metrics into an overall metric; and

determining whether the unknown object is included in the target set based on the overall metric.