Note

This is the documentation for the latest development branch and may refer to features that are not available in released versions. If you are looking for the documentation for a specific release, use the drop-down menu on the left and select the desired version.

Feature Detection Routine Explanation

Overview

In OpenMV, find_hog, find_lines, find_rects, find_features, and get_regression are some of the functions used for image processing and feature detection.

CanMV supports OpenMV algorithms and can also use these functions.

Function Description

find_hog

The find_hog function is used to detect objects in an image using HOG (Histogram of Oriented Gradients) feature descriptors. HOG is a feature description method commonly used for object detection.

• Syntax

objects = img.find_hog(roi=None, threshold=0.5, min_size=(0, 0))

• Parameter Explanation

  • roi: Optional parameter that defines a Region of Interest. The default value is None, which means detection is performed on the entire image.

  • threshold: Confidence threshold. Only detection results with confidence greater than this value will be returned. The default value is 0.5.

  • min_size: The minimum size of the detected object. Specified as a tuple of (width, height). The default value is (0, 0), which means there is no minimum size limit.

• Return Value

Returns a list containing the detected objects, where each object is a Rect object containing the position and size of the object.

find_lines

The find_lines function is used to detect straight lines in an image. This function is suitable for finding long straight lines in an image.

• Syntax

lines = img.find_lines(threshold=1000, theta_margin=20, rho_margin=20)

• Parameter Explanation

  • threshold: The minimum length of a line. The detected line must exceed this length. The default value is 1000.

  • theta_margin: Tolerance for the angle. The unit is degrees. The default value is 20.

  • rho_margin: Tolerance for the distance. The unit is pixels. The default value is 20.

• Return Value

  Returns a list containing line information. Each line is a Line object containing the start point, end point, length, and angle of the line.

find_rects

The find_rects function is used to detect rectangular regions in an image. This function can be used to find square or rectangular objects in an image.

• Syntax

rects = img.find_rects(threshold=2000, margin=10)

• Parameter Explanation

  • threshold: The minimum area of the rectangle. The default value is 2000.

  • margin: The margin tolerance for rectangle detection. The default value is 10.

• Return Value

  Returns a list containing rectangle information. Each rectangle is a Rect object containing the coordinates and size of the rectangle.

find_features

The find_features function is used to detect feature points in an image. These feature points can be used for tasks such as image matching and tracking.

• Syntax

features = img.find_features(algorithm, threshold=10)

• Parameter Explanation

  • algorithm: The feature detection algorithm to use. Common algorithms include image.FORBIDDEN, image.BRIEF, etc.

  • threshold: The threshold for feature detection. The default value is 10.

• Return Value

  Returns a list containing feature point information. Each feature point is a Feature object containing the position of the feature point and other information.

get_regression

The get_regression function is used to detect regression lines from an image, i.e., straight lines that fit the data points. It is typically used to mark the trend or direction of data.

• Syntax

line = img.get_regression(threshold=1000, min_length=10, max_distance=5)

• Parameter Explanation

  • threshold: The minimum length of the line. The default value is 1000.

  • min_length: The minimum length of the fitted line. The default value is 10.

  • max_distance: The maximum distance allowed from a data point to the regression line. The default value is 5.

• Return Value

  Returns a Line object containing the regression line information, representing the fitted line.

Summary

These functions are important tools in the OpenMV image processing library, each used for different vision tasks:

• find_hog: Detects objects based on HOG features.

• find_lines: Detects straight lines in an image.

• find_rects: Detects rectangular regions in an image.

• find_features: Detects feature points in an image.

• get_regression: Detects regression lines in an image.

Example

Only a demo for finding line segments is listed here. For specific demos, please refer to the routines in the virtual USB drive that comes with the firmware.

# 直线检测示例
#
# 这个示例展示了如何在图像中找到直线。对于在图像中找到的每个直线对象，
# 将返回一个包含该直线旋转的直线对象。

# 注意：直线检测是通过霍夫变换完成的：
# http://en.wikipedia.org/wiki/Hough_transform
# 请阅读上述链接以获取有关 `theta` 和 `rho` 的更多信息。

# find_lines() 找到的是无限长的直线。使用 find_line_segments() 可以找到非无限长的直线。
import time, os, gc, sys

from media.sensor import *
from media.display import *
from media.media import *

DETECT_WIDTH = ALIGN_UP(640, 16)
DETECT_HEIGHT = 480

# 所有的直线对象都有一个 `theta()` 方法来获取其旋转角度（以度为单位）。
# 你可以根据其旋转角度来过滤直线。

min_degree = 0
max_degree = 179

# 所有的直线对象还有 `x1()`, `y1()`, `x2()`, 和 `y2()` 方法来获取它们的端点，
# 以及一个 `line()` 方法来将上述所有值作为一个包含 4 个值的元组返回给 `draw_line()` 使用。

# 关于负的 rho 值：
#
# 一个 [theta+0:-rho] 元组与 [theta+180:+rho] 是一样的。
sensor = None

def camera_init():
    global sensor

    # 使用默认配置构造一个Sensor对象
    sensor = Sensor(width=DETECT_WIDTH,height=DETECT_HEIGHT)
    # sensor复位
    sensor.reset()
    # 设置水平镜像
    # sensor.set_hmirror(False)
    # 设置垂直翻转
    # sensor.set_vflip(False)

    # 设置通道 0 输出大小
    sensor.set_framesize(width=DETECT_WIDTH,height=DETECT_HEIGHT)
    # 设置通道 0 输出格式
    sensor.set_pixformat(Sensor.GRAYSCALE)
    # 使用 IDE 作为显示输出，如果您选择的屏幕无法点亮，请参考API文档中的k230_canmv_display_module_api_manual自行配置
    Display.init(Display.VIRT, width= DETECT_WIDTH, height = DETECT_HEIGHT,fps=100,to_ide = True)
    # 初始化媒体管理器
    MediaManager.init()
    # sensor开始运行
    sensor.run()

def camera_deinit():
    global sensor
    # sensor停止运行
    sensor.stop()
    # 销毁显示
    Display.deinit()
    # 休眠
    os.exitpoint(os.EXITPOINT_ENABLE_SLEEP)
    time.sleep_ms(100)
    # 释放媒体缓冲区
    MediaManager.deinit()

def capture_picture():

    fps = time.clock()
    while True:
        fps.tick()
        try:
            os.exitpoint()
            global sensor
            img = sensor.snapshot()

            # `threshold` 控制在图像中找到的直线数量。只有边缘差异幅度和大于 `threshold` 的直线才会被检测到。

            # 关于 `threshold` 的更多信息 - 图像中的每个像素都会为直线贡献一个幅度值。
            # 所有贡献的总和就是该直线的幅度。然后，当直线合并时，它们的幅度会被相加。
            # 注意，`threshold` 会在合并前过滤掉低幅度的直线。要查看未合并直线的幅度，请将 `theta_margin` 和 `rho_margin` 设置为 0。

            # `theta_margin` 和 `rho_margin` 控制合并相似的直线。如果两条直线的 theta 和 rho 值差异小于该边界，它们会被合并。

            for l in img.find_lines(threshold = 1000, theta_margin = 25, rho_margin = 25):
                if (min_degree <= l.theta()) and (l.theta() <= max_degree):
                    img.draw_line([v for v in l.line()], color = (255, 0, 0))
                    print(l)

            # 将结果绘制到屏幕上
            Display.show_image(img)
            img = None

            gc.collect()
            #print(fps.fps())
        except KeyboardInterrupt as e:
            print("user stop: ", e)
            break
        except BaseException as e:
            print(f"Exception {e}")
            break

def main():
    os.exitpoint(os.EXITPOINT_ENABLE)
    camera_is_init = False
    try:
        print("camera init")
        camera_init()
        camera_is_init = True
        print("camera capture")
        capture_picture()
    except Exception as e:
        print(f"Exception {e}")
    finally:
        if camera_is_init:
            print("camera deinit")
            camera_deinit()

if __name__ == "__main__":
    main()

Tip

For specific interface definitions, please refer to image