Apollo 10.0
自动驾驶开放平台
|
After the previous projection in the preprocessing stage, a projection frame is obtained on the picture, but the obtained projection frame is not completely reliable, so a larger region of interest (Region of Interest ROI) to be calculated by the projected signal light position is used. to determine the exact bounding box of the semaphore. Signal light detection (detect) is a conventional convolutional neural network detection task, which receives images with ROI information as input data and sequentially outputs bounding boxes.
apollo::perception::onboard::TrafficLightDetectComponent
Name | Type | Description | Input channal |
---|---|---|---|
frame | apollo::perception::onboard::TrafficDetectMessage | trafficlight message | /perception/inner/Detection |
>Note: The input channel is structure type data. The default trigger channel is /perception/inner/Detection
. The detailed input channel information is in modules/perception/traffic_light_detection/dag/traffic_light_detection.dag
file. By default, the upstream components of the messages received by the component include traffic_light_region_proposal
.
Name | Type | Description | Output channal |
---|---|---|---|
frame | apollo::perception::onboard::TrafficDetectMessage | trafficlight message | /perception/inner/Retection |
>Note: The output channel is structure type data. The message is defined in the modules/perception/common/onboard/inner_component_messages/traffic_inner_component_messages.h
file. The output channel message data can be subscribed by components in the same process. The detailed output channel information is in modules/perception/traffic_light_detection/conf/traffic_light_detection_config.pb.txt
file.
如果您在使用文档的过程中,遇到任何问题,请到我们在【开发者社区】建立的 反馈意见收集问答页面,反馈相关的问题。我们会根据反馈意见对文档进行迭代优化。