Calculate velocity from optic flow. More...

#include "std.h"
#include "inter_thread_data.h"
#include "lib/vision/image.h"
#include "lib/v4l/v4l2.h"

Include dependency graph for opticflow_calculator.h:

This graph shows which files directly or indirectly include this file:

Data Structures
struct	opticflow_t

Macros
#define	FAST9_MAX_CORNERS 512

Functions
void	opticflow_calc_init (struct opticflow_t opticflow[])
	Initialize the opticflow calculator. More...

bool	opticflow_calc_frame (struct opticflow_t opticflow, struct image_t img, struct opticflow_result_t *result)
	Run the optical flow on a new image frame. More...

bool	calc_fast9_lukas_kanade (struct opticflow_t opticflow, struct image_t img, struct opticflow_result_t *result)
	Run the optical flow with fast9 and lukaskanade on a new image frame. More...

bool	calc_edgeflow_tot (struct opticflow_t opticflow, struct image_t img, struct opticflow_result_t *result)
	Run the optical flow with EDGEFLOW on a new image frame. More...

void	kalman_filter_opticflow_velocity (float velocity_x, float velocity_y, float acceleration_measurement, float fps, float measurement_noise, float process_noise, bool reinitialize_kalman)

Detailed Description

Calculate velocity from optic flow.

Using images from a vertical camera and IMU sensor data.

Definition in file opticflow_calculator.h.

Data Structure Documentation

◆ opticflow_t

struct opticflow_t

Definition at line 40 of file opticflow_calculator.h.

Collaboration diagram for opticflow_t:

Data Fields
int	actfast_gradient_method	Whether to use a simple or Sobel filter.
float	actfast_long_step	Step size to take when there is no texture.
int	actfast_min_gradient	Threshold that decides when there is sufficient texture for edge following.
float	actfast_short_step	Step size to take when there is an edge to be followed.
const struct video_config_t *	camera
uint8_t	corner_method	Method to use for determining where the corners are.
bool	derotation	Derotation switched on or off (depended on the quality of the gyroscope measurement)
float	derotation_correction_factor_x	Correction factor for derotation in x axis, determined from a fit from the gyros and flow rotation. (wrong FOV, camera not in center)
float	derotation_correction_factor_y	Correction factor for derotation in Y axis, determined from a fit from the gyros and flow rotation. (wrong FOV, camera not in center)
bool	fast9_adaptive	Whether the FAST9 threshold should be adaptive.
uint16_t	fast9_min_distance	Minimum distance in pixels between corners.
uint8_t	fast9_num_regions	The number of regions of interest the image is split into.
uint16_t	fast9_padding	Padding used in FAST9 detector.
bool	fast9_region_detect	Decides whether to detect fast9 corners in specific regions of interest or the whole image (only for feature management)
struct point_t *	fast9_ret_corners	Corners.
uint16_t	fast9_rsize	Amount of corners allocated.
uint8_t	fast9_threshold	FAST9 corner detection threshold.
bool	feature_management	Decides whether to keep track corners in memory for the next frame instead of re-detecting every time.
uint16_t	fps
bool	got_first_img	If we got a image to work with.
uint8_t	id
struct image_t	img_gray	Current gray image frame.
bool	just_switched_method	Boolean to check if methods has been switched (for reinitialization)
uint8_t	max_iterations	The maximum amount of iterations the Lucas Kanade algorithm should do.
uint16_t	max_track_corners	Maximum amount of corners Lucas Kanade should track.
bool	median_filter	Decides to use a median filter on the velocity.
uint8_t	method	Method to use to calculate the optical flow.
struct image_t	prev_img_gray	Previous gray image frame.
uint8_t	pyramid_level	Number of pyramid levels used in Lucas Kanade algorithm (0 == no pyramids used)
uint16_t	resolution_factor	The resolution in EdgeFlow to determine the Divergence.
uint16_t	search_distance	Search distance for blockmatching alg.
bool	show_flow	Whether to draw the flow vectors on the image. Watch out! This changes the image as will be received by subsequent processing steps.
uint16_t	subpixel_factor	The amount of subpixels per pixel.
uint8_t	threshold_vec	The threshold in x, y subpixels which the algorithm should stop.
bool	track_back	Whether to track flow vectors back to the previous image, in order to check if the back-tracked flow leads to the original corner position.
uint16_t	window_size	Window size for the blockmatching algorithm (general value for all methods)

Macro Definition Documentation

◆ FAST9_MAX_CORNERS

#define FAST9_MAX_CORNERS 512

Definition at line 88 of file opticflow_calculator.h.

Function Documentation

◆ calc_edgeflow_tot()

bool calc_edgeflow_tot	(	struct opticflow_t *	opticflow,
		struct image_t *	img,
		struct opticflow_result_t *	result
	)

Run the optical flow with EDGEFLOW on a new image frame.

Parameters

[in]	*opticflow	The opticalflow structure that keeps track of previous images
[in]	*state	The state of the drone
[in]	*img	The image frame to calculate the optical flow from
[out]	*result	The optical flow result
	computation	successful

Definition at line 1017 of file opticflow_calculator.c.

Referenced by opticflow_calc_frame().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ calc_fast9_lukas_kanade()

bool calc_fast9_lukas_kanade	(	struct opticflow_t *	opticflow,
		struct image_t *	img,
		struct opticflow_result_t *	result
	)

Run the optical flow with fast9 and lukaskanade on a new image frame.

Parameters

[in]	*opticflow	The opticalflow structure that keeps track of previous images
[in]	*state	The state of the drone
[in]	*img	The image frame to calculate the optical flow from
[out]	*result	The optical flow result

Returns: Was optical flow successful

Definition at line 494 of file opticflow_calculator.c.

References act_fast(), ACT_FAST, opticflow_t::actfast_gradient_method, opticflow_t::actfast_long_step, opticflow_t::actfast_min_gradient, opticflow_t::actfast_short_step, agl_dist_value_filtered, analyze_flow_field(), analyze_linear_flow_field(), bottom_camera, opticflow_t::camera, opticflow_result_t::camera_id, video_config_t::camera_intrinsics, cmp_flow(), opticflow_result_t::corner_cnt, opticflow_t::corner_method, point_t::count, opticflow_t::derotation, opticflow_t::derotation_correction_factor_x, opticflow_t::derotation_correction_factor_y, video_config_t::dev_name, opticflow_result_t::div_size, opticflow_result_t::divergence, linear_flow_fit_info::divergence, flow_t::error, image_t::eulers, EXHAUSTIVE_FAST, opticflow_t::fast9_adaptive, fast9_detect(), FAST9_HIGH_THRESHOLD, FAST9_LOW_THRESHOLD, opticflow_t::fast9_min_distance, opticflow_t::fast9_padding, opticflow_t::fast9_ret_corners, opticflow_t::fast9_rsize, opticflow_t::fast9_threshold, opticflow_t::feature_management, opticflow_result_t::flow_der_x, opticflow_result_t::flow_der_y, flow_t::flow_x, opticflow_result_t::flow_x, flow_t::flow_y, opticflow_result_t::flow_y, camera_intrinsics_t::focal_x, camera_intrinsics_t::focal_y, opticflow_result_t::fps, get_size_divergence(), opticflow_t::got_first_img, image_t::h, opticflow_t::id, image_copy(), image_create(), IMAGE_GRAYSCALE, image_show_flow_color(), image_show_points(), image_switch(), image_to_grayscale(), opticflow_t::img_gray, InitMedianFilterVect3Float, opticflow_t::just_switched_method, LARGE_FLOW_ERROR, LINEAR_FIT, manage_flow_features(), opticflow_t::max_iterations, opticflow_t::max_track_corners, MEDIAN_DEFAULT_SIZE, opticflow_t::median_filter, n_agents, n_samples, n_time_steps, opticflow_result_t::noise_measurement, opticflow, opticFlowLK(), FloatEulers::phi, flow_t::pos, predict_flow_vectors(), opticflow_t::prev_img_gray, FloatEulers::psi, opticflow_t::pyramid_level, opticflow_t::show_flow, SIZE_DIV, opticflow_t::subpixel_factor, opticflow_result_t::surface_roughness, linear_flow_fit_info::surface_roughness, FloatEulers::theta, opticflow_t::threshold_vec, timeval_diff(), opticflow_t::track_back, opticflow_result_t::tracked_cnt, image_t::ts, UpdateMedianFilterVect3Float, VECT3_ASSIGN, opticflow_result_t::vel_body, opticflow_result_t::vel_cam, vel_filt, image_t::w, opticflow_t::window_size, FloatVect3::x, point_t::x, point_t::x_sub, FloatVect3::y, point_t::y, point_t::y_sub, and FloatVect3::z.

Referenced by opticflow_calc_frame().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ kalman_filter_opticflow_velocity()

void kalman_filter_opticflow_velocity	(	float *	velocity_x,
		float *	velocity_y,
		float *	acceleration_measurement,
		float	fps,
		float *	measurement_noise,
		float	process_noise,
		bool	reinitialize_kalman
	)

◆ opticflow_calc_frame()

bool opticflow_calc_frame	(	struct opticflow_t *	opticflow,
		struct image_t *	img,
		struct opticflow_result_t *	result
	)

Run the optical flow on a new image frame.

Parameters

[in]	*opticflow	The opticalflow structure that keeps track of previous images
[in]	*state	The state of the drone
[in]	*img	The image frame to calculate the optical flow from
[out]	*result	The optical flow result