people_detect

Detects the people with the HOG descriptor and the linear SVM (cv::HOGDescriptor).

opencv_apps	1.11 and later (ROS 1), 3.0.0 and later (ROS 2)
ROS	noetic, ROS One, jazzy, kilted, lyrical and later
OpenCV	any version of the distributions
Group	Object detection, the nodes of which publish `~/detections` (DetectionArrayStamped)

Usage

# ROS 2
ros2 launch opencv_apps people_detect.launch.xml image:=/camera/image_raw debug_view:=false
# ROS 1
roslaunch opencv_apps people_detect.launch image:=/camera/image_raw debug_view:=false

Try it with the demo video

Download the input video of the demo above, publish it as image with image_publisher, and run the node on it, in the terminals:

wget https://raw.githubusercontent.com/ros-perception/opencv_apps/indigo/test/face_detector_withface_test_diamondback.mp4
# terminal 1: publish the video as image in a loop
ros2 run image_publisher image_publisher_node face_detector_withface_test_diamondback.mp4 --ros-args -p publish_rate:=10.0 -r image_raw:=image
# terminal 2: run the node
ros2 launch opencv_apps people_detect.launch.xml image:=image
# terminal 3: view the result
ros2 run rqt_image_view rqt_image_view /people_detect/image

HOG finds the people whose whole body is in the image, of about 64x128 pixels or larger.

In ROS 1:

# terminal 1
rosrun image_publisher image_publisher face_detector_withface_test_diamondback.mp4 _publish_rate:=10 __name:=camera
# terminal 2
roslaunch opencv_apps people_detect.launch image:=camera/image_raw

See Credits of the demo videos for the license of the video.

Subscribed topics

Topic	Type
`image`	`sensor_msgs/msg/Image`

Published topics

Topic	Type	Note
`~/detections` (recommended)	opencv_apps/msg/DetectionArrayStamped	Recommended, opencv_apps 3.0.1 and later. The people, as the other detection nodes: the centers of the rectangles, the label `person` and the score of the HOG detector.
`~/found` (compatibility)	opencv_apps/msg/RectArrayStamped	Kept for compatibility, use `~/detections`. The top-left corners of the rectangles.
`~/image`	`sensor_msgs/msg/Image`

Parameters

See also the common parameters.

Name	Type	Default (range)	Changed at run time	Description
`group_threshold`	int	2 (0 - 100)	yes	Coefficient to regulate the similarity threshold. When detected, some objects can be covered by many rectangles. 0 means not to perform grouping.
`hit_threshold`	double	0.0 (0 - 10)	yes	Threshold for the distance between features and SVM classifying plane.
`padding`	int	32 (0 - 128)	yes	Padding of the detection window in pixels, of cv::HOGDescriptor::detectMultiScale().
`queue_size`	int	3	no, restart the node	Specigy queue_size of input image subscribers
`scale0`	double	1.05 (1 - 100)	yes	Coefficient of the detection window increase.
`win_stride`	int	8 (1 - 100)	yes	Window stride. It must be a multiple of block stride.

Source

src/nodelet/people_detect_nodelet.cpp, launch/people_detect.launch.xml