Problem when try to run with my own bag

[maxwu8416]

hi ~
Thanks for your great works and sharing it !
Recent days i was try to run your optimized VINS-MONO with my own bag file, there shows some problem i still cannot figure it out.
And the problem is when i launch the vins_estimator by the command in terminal

roslaunch vins_estimator realsense_color.launch

it seems nothing error message shows up, and then i play the bag file, there has image in "tracked image" window(/attention_viewer/output), but the image was rgb image not the grayscale image then i think that's why there has no feature point detected on it.

My camera hardware : Intel realsense zr300

Launch file i changed:

pwd/vins_estimator/launch/realsense_color.launch

<launch>
    <arg name="config_path" default = "$(find feature_tracker)/../config/realsense/realsense_color_config.yaml" />
	<arg name="vins_path" default = "$(find feature_tracker)/../" />
    
    <node name="feature_tracker" pkg="feature_tracker" type="feature_tracker" output="log">
        <remap from="feature_tracker/image" to="/zr300_node/color/image_raw" />
        <remap from="feature_tracker/output" to="feature_tracker/feature_img" />
        <param name="config_file" type="string" value="$(find feature_tracker)/config/zr300_color.yaml" />
        <param name="vins_folder" type="string" value="$(arg vins_path)" />
    </node>

    <node name="vins_estimator" pkg="vins_estimator" type="vins_estimator" output="screen">
       <param name="config_file" type="string" value="$(arg config_path)" />
       <param name="vins_folder" type="string" value="$(arg vins_path)" />
    </node>

    <node name="pose_graph" pkg="pose_graph" type="pose_graph" output="screen">
        <param name="config_file" type="string" value="$(arg config_path)" />
        <param name="visualization_shift_x" type="int" value="0" />
        <param name="visualization_shift_y" type="int" value="0" />
        <param name="skip_cnt" type="int" value="0" />
        <param name="skip_dis" type="double" value="0" />
    </node>

   <!-- attention visualization -->
    <node name="attention_viewer" pkg="feature_tracker" type="attention_viewer" output="screen">
        <remap from="attention_viewer/image" to="/zr300_node/color/image_raw" />
        <remap from="attention_viewer/feature" to="feature_tracker/feature" />
        <remap from="attention_viewer/selection_info" to="vins_estimator/selection_info" />
        <param name="config_file" type="string" value="$(find feature_tracker)/config/zr300_color.yaml" />
        <param name="vins_folder" type="string" value="$(arg vins_path)" />
    </node>

</launch>

Config file i changed :

pwd/config/realsense/realsense_color_config.yaml

%YAML:1.0

#common parameters
imu_topic: "/zr300_node/imu"
image_topic: "/zr300_node/color/image_raw"
output_path: "/home/max/vins_mit/src/vins_mit"

add the yaml file under feature_tracker's config folder
/pwd/feature_tracker/config/zr300_color.yaml

%YAML:1.0

#camera calibration 
model_type: PINHOLE
camera_name: camera
image_width: 640
image_height: 480
distrotion_parameter:
  k1: 9.2615504465028850e-02
  k2: -1.8082438825995681e-01
  p1: -6.5484100374765971e-04
  p2: -3.5829351558557421e-04
projection_parameter:
  fx: 6.0970550296798035e+02
  fy: 6.0909579671294716e+02
  cx: 3.1916667152289227e+02
  cy: 2.3558360480225772e+02

# Feature Tracker Parameters
stride: 2           # only process every `stride`th frame 
                    
eqalization: 1      # histogram equalization for lighting variance
                    
border_margin: 1    # ignore the features found in border pixels
                   
min_distance: 30    # GFTT minimum distance 
                    
max_feature: 150    # number of GFTT to maintain
                   
F_reproj_error: 1.0 # reprojection error threshold for findFundamentalMat
                   

[plusk01]

Your suspicion about the image being color may be valid. This version has a reimplementation of the vins-mono feature tracker. So there are likely edge cases that I did not account for when writing it since I only tested it on the euroc dataset.