Academia.eduAcademia.edu

Figure from:1. INTRODUCTION Track-to-Track Association Using Attributes

See full PDFdownloadDownload figure
Fig. 1. The probability mass function of the local classifier sufficient statistics (1,0?) under the same class hypothesis (likelihood functio1 of H,). 1

Figure 1 The probability mass function of the local classifier sufficient statistics (1,0?) under the same class hypothesis (likelihood functio1 of H,). 1

Related Figures (5)

where x! and «/ are the true classes of tracks i and j, respectively. Note that while in (4) the total probability theorem was used with the diffuse (or improper) prior (6) to yield (7), in (26) the proper (and not necessarily uniform) prior 1° was used (since « takes values in a finite set). In (26) it is assumed that the classification errors are independent across time and sensors.
where x! and «/ are the true classes of tracks i and j, respectively. Note that while in (4) the total probability theorem was used with the diffuse (or improper) prior (6) to yield (7), in (26) the proper (and not necessarily uniform) prior 1° was used (since « takes values in a finite set). In (26) it is assumed that the classification errors are independent across time and sensors.
1 i 2) space) based on the likelihood function of H, with 5% probability of incorrectly rejecting it. Fig. 2. Decision regions (attribute gate in the (v
1 i 2) space) based on the likelihood function of H, with 5% probability of incorrectly rejecting it. Fig. 2. Decision regions (attribute gate in the (v
Fig. 3. The probability mass function of the approximate local classifier sufficient statistic (v? — vi) and the corresponding moment-matched normal probability distribution under the “same class” hypothesis (likelihood function of H,). The 95% attribute gate for this difference is [—4,4].
Fig. 3. The probability mass function of the approximate local classifier sufficient statistic (v? — vi) and the corresponding moment-matched normal probability distribution under the “same class” hypothesis (likelihood function of H,). The 95% attribute gate for this difference is [—4,4].
Fig. 4. The surface of the log-likelihood ratio between the same class (H,) and different classes (Hp) vs. the local classifier sufficient statistics (v!,v2). Note the two peaks, one for (viv?) both large, one for both small.
Fig. 4. The surface of the log-likelihood ratio between the same class (H,) and different classes (Hp) vs. the local classifier sufficient statistics (v!,v2). Note the two peaks, one for (viv?) both large, one for both small.
ig. 5. Decision region for the likelihood ratio test with 5% pecbabiy of incorrectly rejecting H,. This is the 95% attribute gate in the (1 5 y?) space.
ig. 5. Decision region for the likelihood ratio test with 5% pecbabiy of incorrectly rejecting H,. This is the 95% attribute gate in the (1 5 y?) space.