Object recognition device
An object recognition device ( 3 ) includes: a processor ( 4 ) that inputs point cloud sensor data sensed by a sensor (externality recognition sensor ( 1 )) to a software pipeline formed of a plurality of stages and executes object recognition processing; and a memory. The processor ( 4 ) has a plurality of processing units ( 5 ) that executes processing constituting the object recognition processing allocated to each of stages A to C. The memory has a plurality of task queue regions (task queues A to C) that temporarily stores the output data of each of the stages. In a case where the output data of each of the stages is not valid upon input to the subsequent stage, the processor ( 4 ) rejects the output data without storing same in the respective task queue regions.
1 . An object recognition device comprising:
a processor, wherein the processor
receives point cloud sensor data from a sensor that senses the point cloud sensor data,
inputs the point cloud sensor data to a software pipeline formed of a plurality of stages,
executes object recognition processing based on the point cloud sensor data in the software pipeline, and
has a plurality of processing units that executes respective object recognition processing of each of the stages producing corresponding output data of each of the stages; and
a memory having a plurality of task queue regions that temporarily stores the corresponding output data of each of the stages, wherein in a case when the corresponding output data of each of the stages is not valid upon input to the subsequent stage, the processor rejects the corresponding output data without storing the corresponding output data in a respective task queue region.
2 . The object recognition device according to claim 1 , wherein:
the processor comprises a plurality of enqueue units, and
in the case when the corresponding output data of each of the stages is not valid upon input to the subsequent stage, each enqueue unit rejects the corresponding output data without storing the corresponding output data in the respective task queue region.
3 . The object recognition device according to claim 2 , wherein the object recognition device is mounted on a vehicle and the sensor is a LiDAR sensor.
4 . The object recognition device according to claim 2 , wherein each of the enqueue units stores the corresponding output data in the respective task queue regions only in a case when the corresponding output data of each of the stages is valid upon input to the subsequent stage.
5 . The object recognition device according to claim 2 , wherein:
in a case when observation values cannot be acquired, the sensor includes, in the sensor data, information indicating that the sensor data is invalid, and outputs the sensor data,
in a case when the information indicating that the sensor data is invalid is included in the sensor data, the processing unit of a first stage includes, in the corresponding output data, information indicating that the sensor data is invalid, and outputs the corresponding output data, and
in a case when the information indicating that the sensor data is invalid is included in the output data, the corresponding output data of the first stage is not valid upon input to the subsequent stage.
6 . The object recognition device according to claim 5 , wherein, in a case when the corresponding output data of the first stage is not valid upon input to the subsequent stage, the enqueue unit of the first stage does not store the corresponding output data in the respective task queue region of the first stage.
7 . The object recognition device according to claim 2 , wherein, in a case when the corresponding output data of the respective stage is within a range of a three-dimensional space which the sensor is capable of sensing, the corresponding output data of the stage is valid upon input to the subsequent stage.
8 . The object recognition device according to claim 2 , wherein a second and subsequent output data that is a duplicate among the corresponding output data of the respective stage is not valid.
9 . The object recognition device according to claim 1 , wherein priorities of the task queue regions are based on a number of pieces of the unprocessed output data stored in the task queue regions of each of the stages.
10 . The object recognition device according to claim 9 , wherein the priorities of the task queue regions are based on a number of pieces of unprocessed output data stored in the task queue regions of each of the stages.
11 . The object recognition device according to claim 10 , wherein the plurality of processing units executes the processing of the subsequent stage that processes the output data stored in a highest priority task queue region.
12 . The object recognition device according to claim 11 , wherein the processor executes the object recognition processing by using the software pipeline for each piece of sensor data corresponding to a predetermined range of azimuth angles sensed by the sensor.
13 . The object recognition device according to claim 1 , wherein the object recognition processing includes at least one of:
writing coordinate format data to a task queue region when the coordinate format data is contained in a predetermined space and has a reliability equal to or greater than a fixed value;
writing cluster data to the task queue region when the cluster data is contained in the predetermined space and a number of constituent points of a cluster to which the cluster data belongs is equal to or more than a fixed value;
writing index information to the task queue region when a section of a XY plane associated with the cluster data is within a fixed time interval; and
updating group information based on the index information in at least one of the task queue regions and a label value.
14 . The object recognition device according to claim 1 , wherein:
a plurality of software pipelines is provided,
the sensor data corresponding to a predetermined range of azimuth angles sensed by the sensor is input to each of the software pipelines, and
the processor executes the object recognition processing using the plurality of software pipelines.
15 . The object recognition device according to claim 1 , wherein each processing unit executes processing according to a respective partial recognition algorithm that constitutes the object recognition processing of each of the stages.