CN120431476A - Straw coverage recognition method based on machine vision - Google Patents
Straw coverage recognition method based on machine visionInfo
- Publication number
- CN120431476A CN120431476A CN202510948496.3A CN202510948496A CN120431476A CN 120431476 A CN120431476 A CN 120431476A CN 202510948496 A CN202510948496 A CN 202510948496A CN 120431476 A CN120431476 A CN 120431476A
- Authority
- CN
- China
- Prior art keywords
- straw
- image
- difference
- machine vision
- height difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
本发明涉及图像分析技术领域，具体涉及一种基于机器视觉的秸秆覆盖度识别方法，包括以下步骤：通过偏振多角度成像设备采集农田偏振图像组，所述图像组包括多偏振角下的近红外波段图像；基于所述偏振图像组计算镜面反射抑制图像，对抑制后图像进行视差值计算，并生成秸秆‑土壤的高度差标记图；根据所述高度差标记图统计覆盖度，覆盖度=秸秆标记像素数/土壤暴露区域像素数×100%。本发明，最终的秸秆覆盖度计算结果具有更强稳定性和现场适应性，在多种干扰场景（杂草、水膜、破碎秸秆）下降低误差，为智能农机播种决策提供可信依据。
The present invention relates to the field of image analysis technology, and specifically to a machine vision-based straw cover identification method. The method comprises the following steps: using a polarization multi-angle imaging device to capture a set of polarized images of farmland, the set comprising near-infrared band images at multiple polarization angles; calculating a specular reflection suppression image based on the polarized image set, performing parallax calculations on the suppressed image, and generating a straw-soil height difference marker map; and calculating straw cover based on the height difference marker map, where coverage = number of straw marker pixels / number of exposed soil pixels × 100%. This method achieves a final straw cover calculation with enhanced stability and field adaptability, reducing errors in various interference scenarios (weeds, water film, and broken straw), providing a reliable basis for intelligent agricultural machinery sowing decisions.
Description
Technical Field
The invention relates to the technical field of image analysis, in particular to a straw coverage recognition method based on machine vision.
Background
Straw coverage is one of the important agronomic measures of current protective cultivation and reducing soil wind erosion and water erosion, accurate monitoring of coverage has important significance for guiding no-tillage seeding and evaluating operation quality, and the traditional straw coverage identification method mainly depends on the following technical paths:
The manual photographing and image labeling method is that an operator photographs farmland surface images on the ground by using a common camera, and coverage estimation is carried out by manually labeling straw areas and soil areas. The method is time-consuming, labor-consuming, high in subjectivity and difficult to meet the automatic detection requirement of a large-area farmland.
RGB or multispectral imaging method, which is to classify and identify the straws and soil in the image based on color or spectrum difference. However, under the conditions of strong illumination change, weed coverage or straw rot and fading, color and spectral characteristics are easily confused, so that the recognition accuracy is obviously reduced.
In addition, some researches try to detect the farmland surface by using a polarization imaging mode, but a rotary polaroid and single-camera acquisition mode is adopted, the image acquisition time difference is large (generally more than 200 ms), so that the image dislocation is serious, and blurring and spatial offset are extremely easy to generate particularly when the images are acquired on a mobile agricultural machine, and the method is difficult to be used for subsequent three-dimensional structural analysis.
In summary, the existing method has the following general problems that imaging time-space inconsistency, insufficient polarization data utilization, difficulty in distinguishing weed and straw structures, too rough definition of soil areas and lack of a coverage calculation model with high robustness, and a novel straw coverage recognition method integrating image consistency guarantee, strong light reflection inhibition capability, accurate structural feature expression and definite physical constraint of the calculation model is needed.
Disclosure of Invention
The invention provides a machine vision-based straw coverage recognition method.
A straw coverage recognition method based on machine vision comprises the following steps:
S1, collecting a farmland polarized image group through polarized multi-angle imaging equipment, wherein the image group comprises near infrared band images under multiple polarized angles;
S2, calculating a specular reflection inhibition image based on the polarized image group, calculating a parallax value of the inhibited image, and generating a straw-soil height difference marker map;
s3, counting coverage according to the height difference mark graph, wherein the coverage=the number of straw mark pixels/the number of soil exposure area pixels multiplied by 100%;
in the step S2, the region with the parallax value larger than the preset difference threshold is marked as a straw region.
Optionally, the S1 specifically includes:
s11, splitting an incident light path into three channels by adopting a beam splitter prism, and respectively installing linear polarizers under multiple polarization angles in each channel;
S12, setting a narrow-band filter at a near-infrared band;
S13, driving three CMOS sensors to collect simultaneously through a synchronous exposure controller, and generating a polarized image group with aligned space positions.
Optionally, the band of the near infrared band imageThe reflection difference between the straw and the background soil is enhanced.
Optionally, the multiple polarization angles include polarization angles of 0 °, 45 °, and 90 ° to comprehensively obtain the surface reflection characteristics and to calculate the polarization angle difference value.
The cube type beam splitting prism is adopted to split an incident beam into three channels according to the light path direction, each channel is respectively provided with a linear polaroid, and the polarization directions are respectively as follows:
0 ° (parallel polarization direction);
45 ° (oblique polarization direction);
90 ° (perpendicular polarization direction).
Each surface of the beam-splitting prism is plated with a near infrared band antireflection film, and the average transmittance meets the following conditions: The problems of inconsistent field of view and image dislocation caused by the traditional rotary polaroid are avoided.
The polarization angles of 0 DEG, 45 DEG and 90 DEG can comprehensively obtain the surface reflection characteristics, the 0 DEG and 90 DEG represent two orthogonal directions of polarized light, which is helpful for extracting the difference between specular reflection and diffuse reflection, the 45 DEG direction is added to further capture the middle polarization characteristics caused by asymmetric surfaces or inclined structures (such as straw rising and turning parts), and the three angle combinations can be used for calculating the polarization angle difference values (such as maximum value, minimum value, contrast and the like) to improve the distinguishing capability of reflection modes. Specular reflection is strongest at a specific polarization angle, and can be generally inhibited by taking the minimum value (min) or the degree of polarization (DoP) of images at different polarization angles, wherein only two angles of 0 DEG and 90 DEG may not be enough to cover reflection in certain directions, and the addition of 45 DEG improves the inhibition capability of irregular reflection areas, so that the 'real structure' rather than the 'specular artifact' can be effectively judged under different incident angles.
In the near infrared narrow band filtering arrangement, the method comprises adding a center wavelength ofThe narrow-band filter with the wavelength of nm has the following filtering range: ;
i.e. half-wave width Nm, which is used for inhibiting visible light interference and enhancing the reflectivity difference of the straw and weeds in the near infrared region (the reflectivity of the weeds at 870nm is sharply increased, and the interference is remarkable).
The sensor synchronous exposure control specifically further comprises the step of driving the three CMOS image sensors to simultaneously expose through an external synchronous exposure controller, wherein the exposure starting time error meets the following conditions: in order to ensure that the image acquisition process has almost no time dislocation under the running state of a high-speed mobile platform (agricultural machinery).
For example, vehicle speedKm/h, inThe displacement is only:
。
The wavelength range (850 nm.+ -. 15 nm) is selected according to the following:
1. The visible light interference is avoided, the imaging stability is improved, the visible light wave band (about 400-700 nm) is subjected to strong interference of sunlight intensity change, leaf surface reflection and color difference, the robustness of traditional RGB or broad spectrum imaging in straw and soil identification is poor, the near infrared wave band (NIR) is especially near 850nm, the interference is avoided, and the imaging stability is improved, due to the fact that the illumination fluctuation is relatively small and the reflection is even under natural conditions, the imaging stability is improved, and the imaging method is suitable for identification of texture and structure difference.
The 2.850nm band is a characteristic window in plant identification, in the spectral characteristics of plants, the main absorption band of chlorophyll is positioned in red light and blue light regions (approximately 430nm and 660 nm) and the reflection is obviously enhanced in a near infrared region (700-900 nm), however, straw is taken as dead plant tissue, the cell structure of the straw is degraded, the reflectivity at 850nm is far lower than that of fresh vegetation, the reflection of dry soil is stable in the band, and weeds (particularly living blades) have steep reflectivity (up to 30% reflection) near 870nm and can introduce such interference signals if the filter band is too wide.
3. The measured spectrum data shows:
The reflectivity of the soil is stable at 800-900nm and is about 20-30%;
the reflectivity of the straw is obviously higher than that of soil (the difference is about 10-20%) at 850nm, but is far lower than that of the fresh green plants;
the reflectivity of weeds is rapidly increased at 870-880nm (30% compared with 850 nm), and the high contrast area of straw is obviously misjudged.
The reflection difference between the straw and the background (soil) is enhanced, meanwhile, the interference of a living plant reflection 'climbing area' (860-900 nm) is avoided, the high reflection area of living weeds just appears at 870-900nm, the bandwidth is controlled at 850+/-15 nm, the introduction of a high reflection value above 870nm into polarized image calculation is effectively avoided, and the misjudgment of a specular inhibition image caused by high reflection weeds is avoided.
By means of 850+/-15 nm band limitation, the unstructured reflection influence of different ground features is effectively reduced, texture consistency among polarized angle images is improved, and reflection difference is ensured to be mainly sourced from a straw three-dimensional structure, but not spectrum disturbance.
Optionally, the S2 specifically includes:
S21, carrying out pixel-level registration on the near infrared band images under the polarization angles, establishing a space coordinate mapping relation, and ensuring that the images of the polarization angles reflect the same physical position at the same coordinate;
s22, calculating a polarization angle difference value of each pixel point;
s23, constructing a specular reflection suppression image, namely taking the minimum value of polarization angle difference values of all pixel points as a reference to generate a normalized specular reflection suppression image:
S24, binocular parallax calculation is carried out on the specular reflection suppression image, and a parallax map is generated;
s25, converting the parallax map into a height difference mark map;
S26, marking a continuous area with the height difference of more than 0.5mm in the height difference marking chart as a straw area.
Optionally, the binocular parallax calculation includes using 45 ° polarized images as reference views, and using a semi-global matching algorithm to respectively perform the binocular parallax calculationAndAnd carrying out stereo matching on the images, outputting two parallax images, and carrying out average fusion to generate a final parallax image.
Optionally, the polarization angle difference value is calculated as:
Wherein, the method comprises the steps of, Indicating a polarization angle ofThe image being at coordinatesThe gray value at which the color is to be changed,A polarization angle difference value graph is shown for representing the reflection angle isomerism.
Optionally, the generating of the level difference marker graph includes converting the parallax value into a true physical level difference, and the conversion is expressed as: , wherein, For a baseline distance, i.e. the center distance of the two imaging channels,For the focal length of the imaging lens,For the physical level differences at coordinates (x, y) in the level difference map,Is the disparity value at coordinates (x, y).
Optionally, the step S3 specifically includes:
s31, carrying out morphological closing operation on the height difference mark graph, wherein the closing operation structural element is a round nucleus, and the broken straw area is connected;
S32, extracting a communicating region with a height difference of more than 0.5mm as a straw region, and counting the total number of pixels ;
S33, extracting a continuous area with the height difference less than or equal to 0.2mm as a soil exposure area;
S34, counting total number of pixels in the soil exposure area ;
S35, calculating coverage:。
Optionally, the soil exposure area further satisfies:
The area of the region is more than or equal to 100cm 2, and converted into image pixels which are more than or equal to 500 pixels;
the standard deviation of the height difference inside the area is less than or equal to 0.05mm.
The invention has the beneficial effects that:
according to the invention, by adopting a beam splitting prism and three-channel synchronous exposure architecture and matching with a fixed 0 degree/45 degree/90 degree polaroid and a narrow-band filter with the center wavelength of 850nm +/-15 nm, the problem of image dislocation caused by rotating the polaroid by a traditional single sensor is effectively avoided, meanwhile, spectral noise caused by the reflection surge of living weeds at 870nm is filtered, in addition, the specular reflection inhibition image constructed based on the polarization angle difference value strengthens the expression of straws and soil on the optical structure difference, so that obvious structural texture difference can be formed in a non-contact state of straw areas, and ideal input is provided for subsequent high-level analysis.
According to the invention, a semi-global matching algorithm guided by a polarization diagram is used in farmland straw identification, a 45-degree polarization diagram is used as a reference view, and difference parallax calculation is performed by combining 0-degree view with 90-degree view, effective straw parallax information can still be extracted in a low-texture, partial shielding or overlapping area, and the generated height difference marker diagram has remarkable distinguishing capability under the thickness level of rotten straws in cooperation with the actual calibrated baseline distance and focal length parameters, and the structural texture performance of the height difference marker diagram is superior to that of the traditional RGB or NIR depth mapping method.
According to the invention, a soil exposure area identification mechanism based on three physical constraints of a height difference threshold value, an area and a flatness standard deviation is designed, so that the problem that stones, ruts and shadows are mistakenly identified as bare soil by a traditional method is effectively solved. Wherein, the areas with the height difference less than or equal to 0.2mm, the area more than or equal to 100cm 2 and the standard deviation less than or equal to 0.05mm are screened, so that the extracted soil area has more surface consistency and physical rationality. The final straw coverage calculation result has stronger stability and field adaptability, reduces errors under various interference scenes (weeds, water films and crushed straw), and provides a credible basis for intelligent agricultural machinery sowing decisions.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only of the invention and that other drawings can be obtained from them without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method according to an embodiment of the invention;
fig. 2 is a schematic diagram of a coverage calculation process according to an embodiment of the invention.
Detailed Description
The invention will now be described in more detail with reference to the drawings and specific embodiments thereof, which are intended to be illustrative of the embodiments and not limiting thereof, as other alternatives will be apparent to those skilled in the art.
As shown in fig. 1-2, a machine vision-based straw coverage recognition method includes the following steps:
S1, collecting a farmland polarized image group through polarized multi-angle imaging equipment, wherein the image group comprises near infrared band images under multiple polarized angles;
S2, calculating a specular reflection inhibition image based on the polarized image group, calculating a parallax value of the inhibited image, and generating a straw-soil height difference marker map;
S3, counting coverage according to the height difference mark graph, wherein the coverage=the number of straw mark pixels/the number of soil exposure area pixels multiplied by 100%;
In S2, the region with the parallax value larger than the preset difference threshold is marked as a straw region.
S1 specifically comprises:
s11, designing a beam splitting polarization imaging channel, namely adopting a cube beam splitting prism to split an incident beam into three channels according to the light path direction, and respectively configuring linear polarizers for each channel, wherein the polarization directions are as follows:
0 ° (parallel polarization direction);
45 ° (oblique polarization direction);
90 ° (perpendicular polarization direction).
Each surface of the beam-splitting prism is plated with a near infrared band antireflection film, and the average transmittance meets the following conditions: The problems of inconsistent field of view and image dislocation caused by the traditional rotary polaroid are avoided.
The polarization angles of 0 DEG, 45 DEG and 90 DEG can comprehensively obtain the surface reflection characteristics, the 0 DEG and 90 DEG represent two orthogonal directions of polarized light, which is helpful for extracting the difference between specular reflection and diffuse reflection, the 45 DEG direction is added to further capture the middle polarization characteristics caused by asymmetric surfaces or inclined structures (such as straw rising and turning parts), and the three angle combinations can be used for calculating the polarization angle difference values (such as maximum value, minimum value, contrast and the like) to improve the distinguishing capability of reflection modes. Specular reflection is strongest at a specific polarization angle, and can be generally inhibited by taking the minimum value (min) or the degree of polarization (DoP) of images at different polarization angles, wherein only two angles of 0 DEG and 90 DEG may not be enough to cover reflection in certain directions, and the addition of 45 DEG improves the inhibition capability of irregular reflection areas, so that the 'real structure' rather than the 'specular artifact' can be effectively judged under different incident angles.
S12, near infrared narrow-band filtering is set, namely, a center wavelength is added into each channelThe narrow-band filter with the wavelength of nm has the following filtering range: ;
i.e. half-wave width Nm, which is used for inhibiting visible light interference and enhancing the reflectivity difference of the straw and weeds in the near infrared region (the reflectivity of the weeds at 870nm is sharply increased, and the interference is remarkable).
S13, sensor synchronous exposure control, namely driving the three CMOS image sensors to simultaneously expose through an external synchronous exposure controller, wherein the exposure initial time error meets the following conditions: in order to ensure that the image acquisition process has almost no time dislocation under the running state of a high-speed mobile platform (agricultural machinery).
For example, vehicle speedKm/h, inThe displacement is only:
。
The wavelength range (850 nm.+ -. 15 nm) is selected according to the following:
1. The visible light interference is avoided, the imaging stability is improved, the visible light wave band (about 400-700 nm) is subjected to strong interference of sunlight intensity change, leaf surface reflection and color difference, the robustness of traditional RGB or broad spectrum imaging in straw and soil identification is poor, the near infrared wave band (NIR) is especially near 850nm, the interference is avoided, and the imaging stability is improved, due to the fact that the illumination fluctuation is relatively small and the reflection is even under natural conditions, the imaging stability is improved, and the imaging method is suitable for identification of texture and structure difference.
The 2.850nm band is a characteristic window in plant identification, in the spectral characteristics of plants, the main absorption band of chlorophyll is positioned in red light and blue light regions (approximately 430nm and 660 nm) and the reflection is obviously enhanced in a near infrared region (700-900 nm), however, straw is taken as dead plant tissue, the cell structure of the straw is degraded, the reflectivity at 850nm is far lower than that of fresh vegetation, the reflection of dry soil is stable in the band, and weeds (particularly living blades) have steep reflectivity (up to 30% reflection) near 870nm and can introduce such interference signals if the filter band is too wide.
3. The measured spectrum data shows:
The reflectivity of the soil is stable at 800-900nm and is about 20-30%;
the reflectivity of the straw is obviously higher than that of soil (the difference is about 10-20%) at 850nm, but is far lower than that of the fresh green plants;
the reflectivity of weeds is rapidly increased at 870-880nm (30% compared with 850 nm), and the high contrast area of straw is obviously misjudged.
The reflection difference between the straw and the background (soil) is enhanced, meanwhile, the interference of a living plant reflection 'climbing area' (860-900 nm) is avoided, the high reflection area of living weeds just appears at 870-900nm, the bandwidth is controlled at 850+/-15 nm, the introduction of a high reflection value above 870nm into polarized image calculation is effectively avoided, and the misjudgment of a specular inhibition image caused by high reflection weeds is avoided.
By means of 850+/-15 nm band limitation, the unstructured reflection influence of different ground features is effectively reduced, texture consistency among polarized angle images is improved, and reflection difference is ensured to be mainly sourced from a straw three-dimensional structure, but not spectrum disturbance.
S2 specifically comprises:
S21, pixel level registration processing, namely performing pixel level registration on the near infrared images under the polarization angles of 0 degree, 45 degrees and 90 degrees, establishing a unified space coordinate mapping relation, and ensuring that the images of all the polarization angles are at the same coordinate Where the same physical location is reflected.
S22, calculating the polarization angle difference value, namely calculating the polarization angle difference value for each pixel point:
Wherein, the method comprises the steps of, Indicating a polarization angle ofThe image being at coordinatesThe gray value at which the color is to be changed,A polarization angle difference value graph is shown for representing the reflection angle isomerism.
S23, constructing a specular reflection suppression image, namely constructing a normalized specular reflection suppression image based on the polarization angle difference value graph:
;
wherein, the Representing the minimum value of the difference values of all pixels in the whole image;, Representing the gray value of the specular reflection inhibition image, the range is 0-255, and the dynamic range is ;
The method is used for eliminating the interference of illumination intensity change on image difference under different shooting conditions and improving the significance of the straw structure.
S24, binocular parallax calculation toThe polarized image is taken as a reference view;
Using a Semi-global matching (Semi-GlobalMatching, SGM) algorithm to respectively pair AndPerforming stereo matching on the images;
Outputting two disparity maps 、Average fusion is carried out to generate a final parallax image 。
S25, generating a height difference mark graph, namely converting pixel parallax values into real physical height differences, and using the following formula: , wherein, The value range is 20 cm to 30cm for the baseline distance, namely the center distance of the two imaging channels,For the focal length of the imaging lens, the range of the value is 8-12mm,Is the physical height difference at coordinates (x, y) in the height difference map, in millimeters,Is the disparity value at coordinates (x, y) in pixels.
S26, marking the straw area, namely marking the continuous area meeting the following conditions in the height difference graph as the straw area, ifmm Straw area, wherein the height threshold value of 0.5mm is a robust segmentation standard set based on the measured 382 groups of rotten straw thickness samples and combined with standard deviation (0.12 mm).
S3 specifically comprises:
S31, morphological closing operation processing, namely executing morphological closing operation on the height difference mark graph to eliminate pinholes and connection fracture areas, wherein the morphological closing operation processing comprises the following steps of:
the structural element core adopts a circular morphology core with the size of 5 multiplied by 5;
the operation sequence is expansion-corrosion, and the connection of broken straw areas is realized.
The size of the round core is 4.2mm based on the measured average width of the broken straw, and the diameter of the corresponding pixel under the condition of adapting the spatial resolution is about 5 pixels.
S32, extracting a straw region, namely extracting a connected region meeting the following conditions from a closed operation result: And counting the total number of pixels in the area as:
Wherein, the method comprises the steps of, Representing the value of a pixel point in the level difference map,Represents a set of connected areas with a height difference greater than 0.5mm, namely straw areas,Representing the number of straw marking pixels.
S33, extracting a soil exposure area, wherein the judgment conditions for defining the soil area extraction are as follows:
;
wherein, the Representing a candidate low-height region,The area of the region is represented, in terms of pixels,Representing a regionStandard deviation of the inner height difference.
S34, counting the total number of pixels in the soil exposed area, namely summing the total number of pixels in all areas meeting the triple condition to obtain: Wherein, the method comprises the steps of, Representing the satisfaction of the height differenceMm, areaPixel, standard deviationThe set of regions of mm,Representing the number of pixels in the soil exposed area.
S35, a final coverage calculation formula is as follows:。
The invention is intended to cover any alternatives, modifications, equivalents, and variations that fall within the spirit and scope of the invention. In the following description of preferred embodiments of the invention, specific details are set forth in order to provide a thorough understanding of the invention, and the invention will be fully understood to those skilled in the art without such details. In other instances, well-known methods, procedures, flows, components, circuits, and the like have not been described in detail so as not to unnecessarily obscure aspects of the present invention.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (10)
1. The straw coverage recognition method based on machine vision is characterized by comprising the following steps of:
S1, collecting a farmland polarized image group through polarized multi-angle imaging equipment, wherein the image group comprises near infrared band images under multiple polarized angles;
S2, calculating a specular reflection inhibition image based on the polarized image group, calculating a parallax value of the inhibited image, and generating a straw-soil height difference marker map;
s3, counting coverage according to the height difference mark graph, wherein the coverage=the number of straw mark pixels/the number of soil exposure area pixels multiplied by 100%;
in the step S2, the region with the parallax value larger than the preset difference threshold is marked as a straw region.
2. The machine vision-based straw coverage recognition method according to claim 1, wherein S1 specifically comprises:
s11, splitting an incident light path into three channels by adopting a beam splitter prism, and respectively installing linear polarizers under multiple polarization angles in each channel;
S12, setting a narrow-band filter at a near-infrared band;
S13, driving three CMOS sensors to collect simultaneously through a synchronous exposure controller, and generating a polarized image group with aligned space positions.
3. The machine vision-based straw coverage recognition method as claimed in claim 1, wherein the near infrared band image is of a bandThe reflection difference between the straw and the background soil is enhanced.
4. The machine vision-based straw coverage recognition method according to claim 1, wherein the multi-polarization angle includes polarization angles of 0 °, 45 °, and 90 ° to comprehensively obtain surface reflection characteristics and to calculate polarization angle difference values.
5. The machine vision-based straw coverage recognition method of claim 4, wherein S2 specifically comprises:
S21, carrying out pixel-level registration on the near infrared band images under the polarization angles, establishing a space coordinate mapping relation, and ensuring that the images of the polarization angles reflect the same physical position at the same coordinate;
s22, calculating a polarization angle difference value of each pixel point;
s23, constructing a specular reflection suppression image, namely taking the minimum value of polarization angle difference values of all pixel points as a reference to generate a normalized specular reflection suppression image:
S24, binocular parallax calculation is carried out on the specular reflection suppression image, and a parallax map is generated;
s25, converting the parallax map into a height difference mark map;
S26, marking a continuous area with the height difference of more than 0.5mm in the height difference marking chart as a straw area.
6. The machine vision-based straw coverage recognition method of claim 5, wherein the binocular disparity calculation comprises using 45 ° polarized images as reference views, and using a semi-global matching algorithm to respectively pairAndAnd carrying out stereo matching on the images, outputting two parallax images, and carrying out average fusion to generate a final parallax image.
7. The machine vision-based straw coverage recognition method of claim 5, wherein the polarization angle difference value is calculated as:
Wherein, the method comprises the steps of, Indicating a polarization angle ofThe image being at coordinatesThe gray value at which the color is to be changed,A polarization angle difference value graph is shown for representing the reflection angle isomerism.
8. The machine vision-based straw coverage recognition method of claim 5, wherein the generating of the level difference signature comprises converting the parallax value into a true physical level difference, the conversion being represented as: , wherein, For a baseline distance, i.e. the center distance of the two imaging channels,For the focal length of the imaging lens,For the physical level differences at coordinates (x, y) in the level difference map,Is the disparity value at coordinates (x, y).
9. The machine vision-based straw coverage recognition method according to claim 1, wherein the step S3 specifically comprises:
s31, carrying out morphological closing operation on the height difference mark graph, wherein the closing operation structural element is a round nucleus, and the broken straw area is connected;
S32, extracting a communicating region with a height difference of more than 0.5mm as a straw region, and counting the total number of pixels ;
S33, extracting a continuous area with the height difference less than or equal to 0.2mm as a soil exposure area;
S34, counting total number of pixels in the soil exposure area ;
S35, calculating coverage:。
10. the machine vision-based straw coverage recognition method of claim 9, wherein the soil exposure area further satisfies:
The area of the region is more than or equal to 100cm 2, and converted into image pixels which are more than or equal to 500 pixels;
the standard deviation of the height difference inside the area is less than or equal to 0.05mm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510948496.3A CN120431476A (en) | 2025-07-10 | 2025-07-10 | Straw coverage recognition method based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510948496.3A CN120431476A (en) | 2025-07-10 | 2025-07-10 | Straw coverage recognition method based on machine vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN120431476A true CN120431476A (en) | 2025-08-05 |
Family
ID=96554274
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510948496.3A Pending CN120431476A (en) | 2025-07-10 | 2025-07-10 | Straw coverage recognition method based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120431476A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110222737A1 (en) * | 2008-12-03 | 2011-09-15 | Bernhard Biskup | Method for measuring the growth of leaf disks of plants and apparatus suited therefor |
| CN111121722A (en) * | 2019-12-13 | 2020-05-08 | 南京理工大学 | Binocular three-dimensional imaging method combining laser dot matrix and polarization vision |
| CN116309670A (en) * | 2023-05-06 | 2023-06-23 | 中国林业科学研究院资源信息研究所 | Bush coverage measuring method based on unmanned aerial vehicle |
| CN118225245A (en) * | 2024-04-09 | 2024-06-21 | 清华大学深圳国际研究生院 | A polarization measurement system and method for quadruple camera |
| CN118762294A (en) * | 2024-06-18 | 2024-10-11 | 江苏师范大学 | An accurate inversion method for field corn biomass based on UAV remote sensing |
-
2025
- 2025-07-10 CN CN202510948496.3A patent/CN120431476A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110222737A1 (en) * | 2008-12-03 | 2011-09-15 | Bernhard Biskup | Method for measuring the growth of leaf disks of plants and apparatus suited therefor |
| CN111121722A (en) * | 2019-12-13 | 2020-05-08 | 南京理工大学 | Binocular three-dimensional imaging method combining laser dot matrix and polarization vision |
| CN116309670A (en) * | 2023-05-06 | 2023-06-23 | 中国林业科学研究院资源信息研究所 | Bush coverage measuring method based on unmanned aerial vehicle |
| CN118225245A (en) * | 2024-04-09 | 2024-06-21 | 清华大学深圳国际研究生院 | A polarization measurement system and method for quadruple camera |
| CN118762294A (en) * | 2024-06-18 | 2024-10-11 | 江苏师范大学 | An accurate inversion method for field corn biomass based on UAV remote sensing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kazmi et al. | Indoor and outdoor depth imaging of leaves with time-of-flight and stereo vision sensors: Analysis and comparison | |
| Gershikov et al. | Horizon line detection in marine images: which method to choose? | |
| Paris et al. | A three-dimensional model-based approach to the estimation of the tree top height by fusing low-density LiDAR data and very high resolution optical images | |
| WO2021084907A1 (en) | Image processing device, image processing method, and image processing program | |
| CN109801302A (en) | A kind of ultra-high-tension power transmission line foreign matter detecting method based on binocular vision | |
| CN104050668A (en) | Object recognition method applied to green tea tender shoots and based on binocular vision technology | |
| Malekabadi et al. | Disparity map computation of tree using stereo vision system and effects of canopy shapes and foliage density | |
| CN112200854B (en) | A three-dimensional phenotypic measurement method of leafy vegetables based on video images | |
| Sánchez-González et al. | Basal area and diameter distribution estimation using stereoscopic hemispherical images | |
| CN116612192A (en) | Digital video-based pest and disease damage area target positioning method | |
| Xie et al. | Multi-feature detection of in-field grain lodging for adaptive low-loss control of combine harvesters | |
| CN114119718A (en) | A binocular vision green vegetation matching and localization method fused with color features and edge features | |
| CN111985435B (en) | A method for drone water monitoring and cruising based on machine vision | |
| CN117456385A (en) | Bamboo forest canopy chlorophyll content estimation method based on unmanned aerial vehicle hyperspectral technology | |
| CN120431476A (en) | Straw coverage recognition method based on machine vision | |
| CN114299144A (en) | An accurate positioning method of rice based on machine vision | |
| CN102628689A (en) | Crop coverage automatic online observation method and device | |
| Liu et al. | Development of a proximal machine vision system for off-season weed mapping in broadacre no-tillage fallows | |
| Reddy et al. | A novel method for water and water canal extraction from landsat-8 oli imagery | |
| CN114898072A (en) | A crop water stress detection method based on 3D temperature characteristics | |
| CN113514402A (en) | System and method for predicting chlorophyll content of winter wheat | |
| CN108109172A (en) | A kind of robotic vision system and method based on new vision | |
| Bakhshipour et al. | Recognition of pomegranate on tree and stereoscopic locating of the fruit | |
| CN113639643A (en) | Crop seedling-stage height detection method based on RGB-D depth camera | |
| CN112102419A (en) | Calibration method and system of dual-light imaging equipment and image registration method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |