Tracking apparatus
An observation allocating section exclusively allocates an observation point to each existing tracker in accordance with a distance between observation information acquired by an information acquiring section and observation information indicated by a predictive distribution. A hypothesis generating section generates a hypothesis likelihood and a hypothesis distribution, for each target tracker and for each hypothesis belonging to a hypothesis group including a first hypothesis and a second hypothesis. The first hypothesis is a hypothesis that the observation point is a result of observation of a subject target. The second hypothesis is a hypothesis that the observation point is not a result of observation of the subject target. The hypothesis likelihood is the likelihood of the hypothesis. The hypothesis distribution is the state distribution updated on the assumption that the hypothesis is correct.
1 . A tracking apparatus comprising a central processing unit (CPU) and a memory storing a program, wherein the CPU is configured to execute the program to perform:
acquiring, for each processing cycle, observation information corresponding to each of a plurality of observation points from a sensor observing a state of a target present around a vehicle;
selecting observation points satisfying a preset generation condition from an observation point group including the plurality of observation points for which the observation information is acquired;
generating a tracking information comprising a state distribution probabilistically indicating a state of observation information for each of the selected observation points;
generating a predictive distribution for each of existing tracking information that are tracking information generated up to a last processing cycle, the predictive distribution being a state distribution in which the state of the observation information of the existing tracking information in a current processing cycle is predicted based on the state distribution included by the existing tracking information; and
updating, for each of the existing tracking information, the state distribution included by the existing tracking information using the predictive distribution that is generated and the observation information obtained from the observation point group in the current processing cycle,
wherein updating the state distribution includes:
exclusively allocating the observation point to each of the existing tracking information in accordance with a distance between the observation information that is acquired and the observation information indicated by the predictive distribution; and
generating, for each combination of two of subject tracking information and two of the observation points among a combination group, the subject tracking information being a existing tracking information to which the observation point is allocated, a combination likelihood that is a likelihood of the combination and a combination distribution that is a state distribution updated on an assumption that the combination is correct,
the subject target is a target associated with the subject tracking information.
2 . The tracking apparatus according to claim 1 , wherein the CPU is further configured to execute the program to perform:
generating the state distribution after updating by integrating, by convolution using the combination likelihood, the combination distributions each generated by the combination for a respective one of the combinations belonging to the combination group.
3 . The tracking apparatus according to claim 1 , wherein
allocating the observation point to each of the existing tracking information comprises allocating up to one of the observation points to each of the existing tracking information.
4 . The tracking apparatus according to claim 1 , wherein the CPU is further configured to execute the program to perform:
generating a presence determination value used to determine, for each of the tracking information, whether any target is associated with the tracking information; and
removing the tracking information for which the presence determination value is equal to or smaller than a pruning threshold.
5 . The tracking apparatus according to claim 1 , wherein the CPU is further configured to execute the program to perform:
a search range for each of the existing tracking information, wherein
allocating the observation point to each of the existing tracking information comprises allocating the observation points present within the search range.
6 . The tracking apparatus according to claim 5 , wherein the CPU is further configured to execute the program to perform:
updating the state of the tracking information using all the observation points present within the search range; and
sorting the tracking information in such a manner that the updating is applied to the tracking information including the search range overlapping another search range and that the updating is applied to the tracking information including the search range not overlapping another search range.
7 . The tracking apparatus according to claim 1 , wherein
the state distribution is represented by a mixture Gaussian distribution.
8 . The tracking apparatus according to claim 7 , wherein the CPU is further configured to execute the program to perform:
calculating, for each of a plurality of Gaussian distributions used as components of the mixture Gaussian distribution representing the state distribution, a mixture likelihood that is a likelihood of the Gaussian distribution; and
removing, from the state distribution, the Gaussian distribution for which the mixture likelihood is equal to or smaller than a local maximum value pruning threshold.