Distinguishing target classes in observations from vertically pointing entomological radars

V. Drake

    Research output: Contribution to journalArticle

    17 Citations (Scopus)

    Abstract

    Vertical-beam entomological radars incorporating rotating polarization and a very-narrow-angle conical scan generate information about both the trajectories of the insect targets and their identities. The identity information comprises four parameters, one a measure of target size, two of shape, and the last the wing-beat frequency. Transformations have been developed to convert these to a set of orthogonal variables that make the widths of peaks in the distributions similar. The existence of distinct target classes has been demonstrated from the radar observations themselves, using an 8-month data set. The analysis uses peak finding followed by association of the peaks found for different parameters, and is achieved through specially developed algorithms and procedures. Peaks have been identified in each parameter separately, in pairs of parameters, and in the four parameters together, the three approaches producing consistent results. Mainly single peaks were found in spring and the first month of summer, and also in late autumn, while multiple peaks occurred through most of summer and early autumn. All peaks are broad, so the number of target classes that can be resolved is not large. Classes were recognized by associating similar combinations of peak positions for the four parameters when these occurred on multiple nights. The most commonly occurring classes corresponded approximately to rules developed previously for identifying targets as locusts and moths. Two methods of using the identified classes to partition a target sample, with different purity-versus-yield trade-offs, are demonstrated.
    Original languageEnglish
    Pages (from-to)3811-3835
    Number of pages25
    JournalInternational Journal of Remote Sensing
    Volume37
    Issue number16
    DOIs
    Publication statusPublished - 2016

    Fingerprint

    autumn
    locust
    summer
    moth
    parameter
    polarization
    trajectory
    radar
    insect
    method
    analysis
    distribution

    Cite this

    @article{0542c0fe641f40c6a2957fe8dab4e095,
    title = "Distinguishing target classes in observations from vertically pointing entomological radars",
    abstract = "Vertical-beam entomological radars incorporating rotating polarization and a very-narrow-angle conical scan generate information about both the trajectories of the insect targets and their identities. The identity information comprises four parameters, one a measure of target size, two of shape, and the last the wing-beat frequency. Transformations have been developed to convert these to a set of orthogonal variables that make the widths of peaks in the distributions similar. The existence of distinct target classes has been demonstrated from the radar observations themselves, using an 8-month data set. The analysis uses peak finding followed by association of the peaks found for different parameters, and is achieved through specially developed algorithms and procedures. Peaks have been identified in each parameter separately, in pairs of parameters, and in the four parameters together, the three approaches producing consistent results. Mainly single peaks were found in spring and the first month of summer, and also in late autumn, while multiple peaks occurred through most of summer and early autumn. All peaks are broad, so the number of target classes that can be resolved is not large. Classes were recognized by associating similar combinations of peak positions for the four parameters when these occurred on multiple nights. The most commonly occurring classes corresponded approximately to rules developed previously for identifying targets as locusts and moths. Two methods of using the identified classes to partition a target sample, with different purity-versus-yield trade-offs, are demonstrated.",
    author = "V. Drake",
    year = "2016",
    doi = "10.1080/01431161.2016.1204028",
    language = "English",
    volume = "37",
    pages = "3811--3835",
    journal = "International Joural of Remote Sensing",
    issn = "0143-1161",
    publisher = "Taylor and Francis Ltd.",
    number = "16",

    }

    Distinguishing target classes in observations from vertically pointing entomological radars. / Drake, V.

    In: International Journal of Remote Sensing, Vol. 37, No. 16, 2016, p. 3811-3835.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Distinguishing target classes in observations from vertically pointing entomological radars

    AU - Drake, V.

    PY - 2016

    Y1 - 2016

    N2 - Vertical-beam entomological radars incorporating rotating polarization and a very-narrow-angle conical scan generate information about both the trajectories of the insect targets and their identities. The identity information comprises four parameters, one a measure of target size, two of shape, and the last the wing-beat frequency. Transformations have been developed to convert these to a set of orthogonal variables that make the widths of peaks in the distributions similar. The existence of distinct target classes has been demonstrated from the radar observations themselves, using an 8-month data set. The analysis uses peak finding followed by association of the peaks found for different parameters, and is achieved through specially developed algorithms and procedures. Peaks have been identified in each parameter separately, in pairs of parameters, and in the four parameters together, the three approaches producing consistent results. Mainly single peaks were found in spring and the first month of summer, and also in late autumn, while multiple peaks occurred through most of summer and early autumn. All peaks are broad, so the number of target classes that can be resolved is not large. Classes were recognized by associating similar combinations of peak positions for the four parameters when these occurred on multiple nights. The most commonly occurring classes corresponded approximately to rules developed previously for identifying targets as locusts and moths. Two methods of using the identified classes to partition a target sample, with different purity-versus-yield trade-offs, are demonstrated.

    AB - Vertical-beam entomological radars incorporating rotating polarization and a very-narrow-angle conical scan generate information about both the trajectories of the insect targets and their identities. The identity information comprises four parameters, one a measure of target size, two of shape, and the last the wing-beat frequency. Transformations have been developed to convert these to a set of orthogonal variables that make the widths of peaks in the distributions similar. The existence of distinct target classes has been demonstrated from the radar observations themselves, using an 8-month data set. The analysis uses peak finding followed by association of the peaks found for different parameters, and is achieved through specially developed algorithms and procedures. Peaks have been identified in each parameter separately, in pairs of parameters, and in the four parameters together, the three approaches producing consistent results. Mainly single peaks were found in spring and the first month of summer, and also in late autumn, while multiple peaks occurred through most of summer and early autumn. All peaks are broad, so the number of target classes that can be resolved is not large. Classes were recognized by associating similar combinations of peak positions for the four parameters when these occurred on multiple nights. The most commonly occurring classes corresponded approximately to rules developed previously for identifying targets as locusts and moths. Two methods of using the identified classes to partition a target sample, with different purity-versus-yield trade-offs, are demonstrated.

    U2 - 10.1080/01431161.2016.1204028

    DO - 10.1080/01431161.2016.1204028

    M3 - Article

    VL - 37

    SP - 3811

    EP - 3835

    JO - International Joural of Remote Sensing

    JF - International Joural of Remote Sensing

    SN - 0143-1161

    IS - 16

    ER -