A method and system for precise throwing with an infrared visual throwing device

Abstract

This invention discloses a method and system for precise delivery using an infrared visual delivery device. Addressing the problem of low delivery efficiency caused by the need for drones to hover above a target point for a certain period during delivery, this invention mounts an infrared visual delivery device onto a drone. Image data collected by the device's infrared and visible light lenses is integrated into the drone's binocular vision system. The drone acquires images including the target point in real time via its forward-facing camera. Based on a preset algorithm, feature points of the target point are calculated and compared with feature points in the real-time images acquired by the binocular vision system to determine the target point's location within the system. This allows for the calculation of the drone's current altitude and horizontal distance from the target point. Combined with the drone's flight speed, the delivery time during flight is calculated. By incorporating this positioning-based delivery algorithm, the drone can achieve precise delivery without hovering, significantly improving delivery efficiency.

Description

Technical Field

[0001] This invention belongs to the technical field of emergency material transportation, and in particular relates to a method and system for precise throwing of an infrared visual throwing device. Background Technology

[0002] Currently, the mainstream method involves drones carrying various supplies, traversing complex terrains and harsh environments, and then opening the drop container at a low altitude to drop the supplies upon arrival at the target location. In other words, the drone flies above the target point, hovers, and then the drop device begins to release the items it carries. This method is inefficient, severely impacting the delivery of emergency supplies and consequently hindering disaster relief efforts. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for precise delivery of an infrared visual projectile. By combining a delivery algorithm based on positioning technology, the drone can achieve precise delivery without hovering, which can greatly improve delivery efficiency.

[0004] To solve the above problems, the technical solution of the present invention is as follows:

[0005] A method for precise throwing with an infrared visual throwing device includes:

[0006] The infrared visual launcher is mounted on the drone, and the image data collected by its infrared lens and visible light lens is connected to the drone's binocular vision system.

[0007] The drone acquires images containing the target point in real time through a forward-facing camera and calculates the feature points of the target point according to a preset algorithm.

[0008] The feature points of the target point are compared with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system.

[0009] Based on the target location, calculate the current altitude and horizontal distance between the drone and the target location. Combined with the drone's flight speed, calculate the drop time point during flight.

[0010] According to an embodiment of the present invention, the step of calculating the feature points and descriptors of the target point according to a preset algorithm further includes:

[0011] The SURF algorithm is used to calculate the feature points and feature point descriptors of the target point.

[0012] According to an embodiment of the present invention, calculating the feature points and feature point descriptors of a target point based on the SURF algorithm further includes:

[0013] The target point image is Gaussian smoothed, and a Gaussian scale space is established. The local extrema of the Hessian matrix determinant are calculated to determine the feature point locations of the target point. For any point in the space with scale σ... The Hessian matrix is ​​defined as follows:

[0014]

[0015] In the formula, L xx It is a second-order transverse differential, L xy It is a second-order cross-differentiation, L yy Second-order longitudinal differential;

[0016] The principal direction of the feature point is calculated, a multi-dimensional feature vector is constructed for the feature point, and the multi-dimensional feature vector is normalized to obtain the feature point descriptor, which contains the neighborhood information of the feature point.

[0017] According to an embodiment of the present invention, comparing the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system further includes:

[0018] Based on the feature point neighborhood information contained in the feature point descriptor calculated by the SURF algorithm, the K nearest neighbor method is used to obtain the two potential best matching points of each feature point in the target image in the image acquired in real time by the binocular vision system.

[0019] The best matching point pair is selected by the ratio between the best and second-best distances between the matching points;

[0020] When the ratio between the best and second-best values ​​is less than the feature intensity threshold, the feature point with the best value is taken as the successfully matched feature point and marked in the image of the binocular vision system.

[0021] According to one embodiment of the present invention, calculating the current altitude and horizontal distance between the UAV and the target point, and combining the UAV's flight speed, further includes calculating the drop time point during flight:

[0022] Horizontal distance x = V0x * t, where V0x is the horizontal velocity of the drone at the time of deployment, and t is time;

[0023] Vertical distance y = V0y*t + (1 / 2)*g*t^2, where V0y is the vertical velocity of the drone at the time of deployment, t is time, and g is the acceleration due to gravity.

[0024] When the vertical velocity V0y of the drone is 0, the vertical distance y = (1 / 2) * g * t^2;

[0025] The vertical distance between the drone and the target point remains essentially constant, thus determining the time the dropped object remains in the air:

[0026] The drone's drop point is the horizontal distance from the target point:

[0027] According to one embodiment of the present invention, calculating the current altitude and horizontal distance between the UAV and the target point, and combining the UAV's flight speed, further includes calculating the drop time point during flight:

[0028] Let the air resistance during the fall of the projectile be k. The aerial and horizontal distances of the drone from the target point are as follows:

[0029]

[0030]

[0031] Where x is the horizontal distance, y is the vertical distance, m is the mass of the thrown object, g is the gravitational acceleration, k is the air resistance, v0 is the horizontal velocity of the UAV, and t is the throwing time;

[0032] By selecting a set of data for m, k, and g, the drop time t is obtained, and then the horizontal distance between the UAV and the target point at the time of drop is determined.

[0033] A point-to-point throwing system for an infrared visual throwing device includes:

[0034] An infrared visual projector is mounted on a drone, and the image data collected by its infrared and visible light lenses is connected to the drone's binocular vision system.

[0035] The drone acquires images including the target point in real time through a forward-facing camera, calculates the feature points of the target point according to a preset algorithm, compares the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system, calculates the current altitude and horizontal distance between the drone and the target point based on the location of the target point, and calculates the release time point during the flight process by combining the drone's flight speed.

[0036] According to an embodiment of the present invention, the infrared visual projector includes an upper shell module, a mechanism module and a lower shell module, wherein the mechanism module is fixedly disposed in the lower shell module and the upper shell module is fixedly connected to the lower shell module;

[0037] The mechanism module integrates a heat dissipation unit, a camera unit, a power supply unit, a lens unit, and a padlock unit. The power supply unit is electrically connected to the heat dissipation unit, the camera unit, the lens unit, and the padlock unit.

[0038] The lens unit includes a lens mounting bracket, an infrared lens, and a visible light lens. The infrared lens and the visible light lens are fixed below the lens mounting bracket. The power supply unit, the camera unit, and the heat dissipation unit are stacked sequentially from bottom to top above the lens mounting bracket.

[0039] The padlock unit includes a first padlock assembly and a second padlock assembly, the first padlock assembly and the second padlock assembly being disposed opposite to each other and located on both sides of the lens fixing bracket;

[0040] The first padlock assembly includes an electronic lock and a bracket. The electronic lock is fixedly connected to the power unit via the bracket, and the electronic lock is used to hang a projectile.

[0041] The bottom of the lower shell module has a slot so that the thrown object can be hooked onto the electronic lock.

[0042] According to one embodiment of the present invention, the upper corner of the electronic lock is recessed to form a notch, and a protrusion for triggering the unlocking of the electronic lock is provided on the notch;

[0043] The lower end of the electronic lock has a groove, and a crossbar that is linked to the protrusion is provided in the groove;

[0044] In the locked state, the crossbar passes through the groove; when the protrusion is pressed, it causes the crossbar to retract, thus unlocking by pressing.

[0045] According to one embodiment of the present invention, the electronic lock is a titanium wire electronic lock, which is controlled by a remote control. When unlocking is required, the titanium wire is heated, causing it to contract instantly, triggering the titanium wire associated structure to unlock, thereby realizing aerial throwing.

[0046] Because the present invention adopts the above technical solution, it has the following advantages and positive effects compared with the prior art:

[0047] This invention discloses a method for precise delivery using an infrared visual delivery device. Addressing the issue of low delivery efficiency caused by the need for drones to hover above a target point for a certain period during delivery, this method mounts an infrared visual delivery device onto a drone. Image data collected by the device's infrared and visible light lenses is integrated into the drone's binocular vision system. The drone acquires images including the target point in real-time via its forward-facing camera and calculates feature points of the target point using a preset algorithm. These feature points are then compared with the feature points of the real-time images acquired by the binocular vision system to determine the target point's location within the system. Based on this location, the drone's current altitude and horizontal distance from the target point are calculated, and combined with the drone's flight speed, the delivery time is determined. The device unlocks upon reaching the delivery time, enabling aerial delivery. This delivery algorithm, combined with positioning technology, allows for precise delivery without the need for drone hovering, significantly improving delivery efficiency. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of an infrared visual projector according to an embodiment of the present invention;

[0049] Figure 2 This is a split structure diagram of an infrared visual projector according to an embodiment of the present invention;

[0050] Figure 3 This is an exploded view of the movement module in one embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of an electronic lock according to one embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of a button in one embodiment of the present invention;

[0053] Figure 6 This is a flowchart of a fixed-point throwing method for an infrared visual throwing device according to an embodiment of the present invention.

[0054] Explanation of reference numerals in the attached figures:

[0055] 1: Top shell module; 2: Mechanism module; 201: Infrared lens; 202: Visible light lens; 203: Lens mounting bracket; 204: Electronic lock; 205: Electronic lock bracket; 206: Power board; 207: Camera board bracket; 208: Camera board; 209: Heat dissipation bracket; 210: Fan; 211: Button; 212: Protrusion; 213: Key body; 214: Spring; 215: Snap ring; 3: Bottom shell module. Detailed Implementation

[0056] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a more detailed account of the fixed-point throwing method and system of the infrared visual throwing device proposed in this invention. The advantages and features of this invention will become clearer from the following description and claims.

[0057] This embodiment provides a method for precise delivery of an infrared visual delivery device. Combined with a delivery algorithm based on positioning technology, the drone can achieve accurate delivery without hovering, greatly improving delivery efficiency. This method relies on a precise delivery system, which includes:

[0058] An infrared visual projector is mounted on a drone, and the image data collected by its infrared and visible light lenses is connected to the drone's binocular vision system.

[0059] The drone acquires images including the target point in real time through a forward-facing camera, calculates the feature points of the target point according to a preset algorithm, compares the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system, calculates the current altitude and horizontal distance between the drone and the target point based on the location of the target point, and calculates the release time point during the flight process by combining the drone's flight speed.

[0060] Please refer to the infrared visual projector. Figure 1 and Figure 2 It includes an upper shell module 1, a movement module 2 and a lower shell module 3. The movement module 2 is fixed inside the lower shell module 3, and the upper shell module 1 and the lower shell module 3 are fixedly connected.

[0061] from Figure 2 As can be seen, the upper shell module 1 is a cover, with a power connector installed on top of the cover. The edge of the cover extends downward to form a cavity. A button is installed at each of the four corners of the cavity, which is used to unlock the locking mechanism (electronic lock) of the launcher.

[0062] The lower shell module 3 is a frame that matches the upper shell module 1. A groove is formed at its bottom, corresponding to the electronic lock in the mechanism module 2, allowing the hook of the thrown object to be easily attached to the electronic lock.

[0063] The camera module 2 integrates a heat dissipation unit, a camera unit, a power supply unit, a lens unit, and a padlock unit. The power supply unit is electrically connected to the heat dissipation unit, camera unit, lens unit, and padlock unit. Please refer to [link / reference]. Figure 3 The power supply unit is a power board 206. A lens mounting bracket 203 is fixed below the power board 206, and an infrared lens 201 and a visible light lens 202 are respectively fixed on the lens mounting bracket 203. A camera board bracket 207, a camera board 208, a heat dissipation bracket 209, and a fan 210 are stacked sequentially above the power board 206 from bottom to top. Multiple electronic locks 204 are connected to both sides of the power board 206, and each pair of electronic locks 204 is fixedly connected by an electronic lock bracket 205.

[0064] The internal unit modules of the entire mechanism module 2 are stacked compactly, making reasonable use of the overall space. The visible light lens, infrared lens and electronic lock are cleverly combined with a simple bracket, making the parabola small in size and light in weight.

[0065] While maintaining a compact overall structure and small size, this embodiment uses a lens mounting bracket 203 to fix the infrared lens 201 and the visible light lens 202, so that the objective lens of the infrared lens 201 is flush with the objective lens of the visible light lens 202 in the installed state, and the two structures are compact.

[0066] Because the infrared lens 201 and the visible light lens 202 are at different heights, the brackets on the lens mounting bracket 203 used to fix the infrared lens 201 and the visible light lens 202 are slightly different. Specifically, the lens mounting bracket 203 includes a first mounting bracket and a second mounting bracket, which are connected left and right. The first mounting bracket is used to fix the infrared lens 201, and the second mounting bracket is used to fix the visible light lens 202. The shape of the first mounting bracket matches the mounting end of the infrared lens 201, and the second mounting bracket is a column-type double-layer bracket. The visible light lens 202 is mounted on the lower bracket to compensate for the height difference between the visible light lens 202 and the infrared lens 201.

[0067] The electronic lock 204 can be electrically unlocked, mechanically unlocked, or a combination of both.

[0068] When the electronic lock 204 is electrically unlocked, a titanium wire electronic lock can be used. It can be operated by remote control. When unlocking is required, the titanium wire is heated, causing it to contract instantly and triggering the associated structure to unlock, thus enabling aerial throwing.

[0069] When electronic lock 204 is mechanically unlocked, the mechanical switch of the electronic lock can be triggered by a button. Please refer to [link / reference]. Figure 4 The electronic lock 204 has a notch formed by a recess at its upper corner, and a protrusion 212 for triggering the lock to unlock is provided on the notch. A groove is formed at the lower end of the electronic lock 204, and a crossbar that is linked to the protrusion 212 is installed within the groove. In the locked state, the crossbar passes through the groove; when the protrusion 212 is pressed, it causes the crossbar to retract, thus unlocking the device. The upper end of the electronic lock 204 is fixedly connected to the upper shell module 1. The cover of the upper shell module 1 has a downward-facing cover edge, and a button 211 is fixed on the corner of the cover edge. In the installed state, the button 211 is opposite to the protrusion 212; when the button 211 is pressed, it squeezes the protrusion, thus unlocking the device. When the thrower or related equipment malfunctions, the button can be manually pressed to unlock the device, allowing for quick replacement of the thrown object.

[0070] Please refer to Figure 5 The button 211 includes a button body 213, a spring 214 and a retaining ring 215. A post is provided on the reverse side of the button body 213 for mounting the spring 214, and the retaining ring 215 is used to fix the spring 214 to the post.

[0071] Compared to existing throwers that use probes to trigger a mechanical switch on the lock body to change the thrown object, which have drawbacks such as low efficiency, the need to locate the probe and align it with the hole to complete the operation, the requirement to use the probe and the risk of losing it, and the potential risk of accidental injury to the human body and the machine, the button-type mechanical unlocking in this embodiment has the following advantages:

[0072] 1. Quick and easy replacement of projectiles by pressing the touch switch at any time;

[0073] 2. Convenient, no need to use other tools to operate;

[0074] 3. Safe and poses no risk to the body or human body.

[0075] Furthermore, the heat dissipation unit in this embodiment adopts a structure of natural ventilation or fan cooling, which can avoid the failure of the core module due to excessive operating temperature and extend its service life.

[0076] In practical applications, when the drone carrying the dropper is in flight, the remote controller communicates with the camera unit to receive video data transmitted by the camera unit and display the images, so that the operator can keep track of the movement of the transported materials in real time and realize visual drop.

[0077] To enable precise delivery of drones without hovering, this embodiment provides a point-to-point delivery method using an infrared visual delivery device, applied to the aforementioned point-to-point delivery system. Please refer to... Figure 6 The method for precisely launching the infrared visual projectile includes the following steps:

[0078] S1: The infrared visual launcher is mounted on the drone, and the image data collected by its infrared lens and visible light lens is connected to the drone's binocular vision system.

[0079] S2: The drone acquires images including the target point in real time through a forward-facing camera and calculates the feature points of the target point according to a preset algorithm;

[0080] S3: Compare the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system.

[0081] S4: Based on the target point's location, calculate the current altitude and horizontal distance between the drone and the target point, and combine this with the drone's flight speed to calculate the drop time during flight.

[0082] Most existing droppers are non-visual (i.e., blind) droppers, while some are visual but lack precise positioning and infrared functionality, making 24 / 7 operation impossible. Drones equipped with such droppers can only operate during the day, and their drop efficiency and accuracy are poor, thus affecting the progress of emergency relief efforts.

[0083] To address this issue, this embodiment employs the aforementioned infrared visual projector, which provides clear images regardless of whether it is day or night. Drones equipped with this infrared visual projector can operate 24 / 7.

[0084] Mounting an infrared visual projector onto a drone can be achieved using conventional methods. It's important to note that the images captured by the infrared and visible light lenses of the projector must be integrated into the drone's binocular vision system. This binocular vision system, located directly in front of the drone, uses two cameras and the principle of binocular vision to obtain depth maps, 3D point clouds, and other information in front of the drone.

[0085] During the targeted drop, the UAV acquires images including the target point in real time through a forward-facing camera and calculates the feature points of the target point according to a preset algorithm. The feature points of the target point are compared with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system. Based on the location of the target point, the UAV calculates the current altitude and horizontal distance between the UAV and the target point, and combines the flight speed of the UAV to calculate the drop time point during the flight.

[0086] Specifically, feature points and feature point descriptors of the target point can be calculated based on the SURF algorithm. According to the feature point neighborhood information contained in the feature point descriptors calculated by the SURF algorithm, the K nearest neighbor method is used to obtain the two potential best matching points of each feature point in the target image in the real-time image acquired by the binocular vision system. The best matching point pair is selected by the ratio between the best and second-best distance between the matching points. When the ratio between the best and second-best distances is less than the feature intensity threshold, the feature point with the best value is taken as the successfully matched feature point and marked in the image of the binocular vision system, thereby determining the location of the target point in the binocular vision system.

[0087] When calculating feature points, the target point image is Gaussian smoothed, and a Gaussian scale space is established. The local extrema of the Hessian matrix determinant are calculated to determine the feature point location of the target point. For any point in the space with scale σ... The Hessian matrix is ​​defined as follows:

[0088]

[0089] In the formula, L xx It is a second-order transverse differential, L xy It is a second-order cross-differentiation, L yy Second-order longitudinal differential. L xx It is the second derivative of Gauss. The result of convolution with I = (x, y), where L xy and L yy similar.

[0090] To maintain rotation invariance of feature points, the principal direction of each feature point is calculated after its position is determined. To this end, within a circular region centered on the feature point and with a radius of 6σ (σ being the scale of the feature point), the image is processed using Haar wavelet (with a side length of 4σ) responses in the x and y directions. A Gaussian weighting function with a scale of 2σ is used to Gaussian weight the Haar wavelet response values, ensuring that responses closer to the feature point contribute more. Then, a sector of size π / 3 is used to traverse the entire circular region, and the direction of the maximum value of the vector sum and magnitude of the horizontal and vertical responses within the domain is defined as the principal direction of the feature point.

[0091] Then, a square window region with one side perpendicular to the main direction and a side length of 20σ is constructed, centered on the feature point. This window region is divided into 4×4 sub-regions. Within each sub-region, the horizontal and vertical Haar wavelet responses of 5σ×5σ sampling points are calculated, denoted as dx and dy respectively. Similarly, a Gaussian weighting function with a scale of 2σ is used to Gaussian weight the Haar wavelet response values ​​to increase robustness to geometric transformations. Then, the response values ​​and their absolute values ​​for each sub-region are summed to form ∑dx, ∑dy, ∑|dx|, ∑|dy|. Thus, each sub-region forms a four-dimensional feature description vector V.

[0092] V=(Σdx,Σdy,Σ|dx|,Σ|dy|)

[0093] For each feature point, a 4×4×4=64-dimensional feature vector is formed.

[0094] To ensure invariance to illumination, the feature vectors are normalized to obtain the final SURF feature point descriptor, which contains neighborhood information of the feature points.

[0095] During feature point matching, based on the feature point neighborhood information contained in the SURF feature point descriptor, the K-nearest neighbor method is used to find the two potential best matching points for each feature point in the target image in the real-time image acquired by the binocular vision system. The best matching point pair is selected by the ratio of the best to the second best distance between the matching points. Only when the ratio of the best to the second best distance is less than the feature intensity threshold α can the feature point with the best value be considered a successfully matched feature point. The feature intensity threshold α is a value between (0,1), which can be determined according to the actual matching situation.

[0096] The k-Nearest Neighbor (KNN) classification algorithm is a theoretically mature method and one of the simplest machine learning algorithms. The idea behind this method is: if a sample's k most similar (i.e., nearest) samples in the feature space mostly belong to a certain class, then that sample also belongs to that class. A detailed description of the KNN method is not provided here.

[0097] Here, the ratio d1 / d2 of the Euclidean distance d1 between the feature vectors of the feature points with the highest similarity to those with the second highest similarity is defined as the feature strength of that feature point. The higher the similarity of a feature point, the smaller its Euclidean distance, and therefore the smaller the ratio of its Euclidean distance to that of the second most similar feature point – the feature strength value.

[0098] The above method calculates the feature points and feature point descriptors of the target point based on the SURF algorithm. The neighborhood information of the feature points contained in the feature point descriptors obtained by the SURF algorithm is then used to match the real-time image to obtain the localization of the target point in the UAV's binocular vision system. Of course, this can also be achieved using other algorithms. For example:

[0099] The drone acquires a target image containing the target point using its front-facing camera, extracts N sub-images from the target image, and records the pixel coordinates of these N sub-images within the target image, forming a sub-image group. The target image can be a visible light image.

[0100] Edge detection is performed on each sub-image in the sub-image group and the real-time image acquired by the binocular vision system. SURF feature points are extracted, and bidirectional two-level matching is performed on the SURF feature points to obtain precise matching points based on spatial domain matching. The real-time image is either an infrared real-time image or a visible light real-time image acquired by the infrared visual projector. The bidirectional two-level matching here includes bidirectional nearest neighbor / second nearest neighbor matching and a random sampling consensus algorithm.

[0101] NSCT transformation is performed on each sub-image and real-time image in the sub-image group to extract low-frequency sub-band images and high-frequency sub-band images respectively. SURF feature point matching is performed on the low-frequency sub-band images, and normalized cross-correlation matching is performed on the high-frequency sub-band images to obtain matching points in the sub-images corresponding to the matching points in the low-frequency sub-band images and the matching points in the high-frequency sub-band images. Duplicate matching points are removed to obtain accurate matching points based on transform domain matching.

[0102] Repeated matching points in the same sub-image are removed based on the exact matching points in the spatial domain and transform domain. Then, according to the pixel coordinate relationship between N sub-images, matching points that do not satisfy the positional relationship between matching points in the sub-images are removed, thus obtaining the matching points of all sub-images.

[0103] The matching points of all sub-images are converted into matching points of the target image. Based on the matching point pairs between the target image and the real-time image, the affine transformation matrix between the target image and the real-time image is calculated to obtain the coordinates of the target point in the target image in the real-time image, thus completing the target recognition.

[0104] This method is applicable to heterogeneous matching applications such as visible light and infrared images, and solves the problem of feature differences between heterogeneous images. Through multi-feature matching, it has the advantages of strong robustness, low probability of mismatch, and high probability of target recognition.

[0105] By obtaining the target point's location in binocular vision, the relative position between the target point and the current drone can be determined. Based on the target point's location, the altitude and horizontal distances between the current drone and the target point are calculated. Combined with the drone's flight speed, the release time during flight is calculated. Specifically:

[0106] Horizontal distance x = V0x * t, where V0x is the horizontal velocity of the drone at the time of deployment, and t is time;

[0107] Vertical distance y = V0y*t + (1 / 2)*g*t^2, where V0y is the vertical velocity of the drone at the time of deployment, t is time, and g is the acceleration due to gravity.

[0108] When the vertical velocity V0y of the drone is 0, the vertical distance y = (1 / 2) * g * t^2;

[0109] The vertical distance between the drone and the target point remains essentially constant, thus determining the time the dropped object remains in the air:

[0110] The drone's drop point is the horizontal distance from the target point: The drone will fly another x distance and then immediately perform an aerial drop.

[0111] The above calculation method does not take air resistance into account. To better reflect reality, air resistance is added to the calculation of the casting time, as follows:

[0112] Let the air resistance during the fall of the projectile be k. The aerial and horizontal distances of the drone from the target point are as follows:

[0113]

[0114]

[0115] Where x is the horizontal distance, y is the vertical distance, m is the mass of the thrown object, g is the gravitational acceleration, k is the air resistance, v0 is the horizontal velocity of the UAV, and t is the throwing time;

[0116] By selecting a set of data for m, k, and g, the release time t is obtained, and thus the horizontal distance from the target point to the drone at the time of release is determined. For example, m is taken as 1 kg, air resistance k as 0.1, and g as 10 m / s^2. It can be deduced that:

[0117] t = -10 * ln((y - 100 * t) / 1000 + 1)

[0118] Therefore, the drop point of the drone is determined to be the horizontal distance from the target point: x = 10 * v0 * (1 - exp(-0.1 * ((y - 1000 * (exp(-0.1t) - 1)) / 100))).

[0119] Different projectiles have different weights and experience different air resistance, therefore, the launching points also differ.

[0120] The infrared visual projector fixed-point delivery method in this embodiment, combined with the delivery algorithm based on positioning technology, allows the drone to achieve precise delivery without hovering, which can greatly improve delivery efficiency.

[0121] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they shall still fall within the protection scope of the present invention.

Claims

1. A method for precise throwing with an infrared visual throwing device, characterized in that, include: The infrared visual launcher is mounted on the drone, and the image data collected by its infrared lens and visible light lens is connected to the drone's binocular vision system. The drone acquires images containing the target point in real time through a forward-facing camera and calculates the feature points of the target point according to a preset algorithm. The feature points of the target point are compared with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system. Based on the target point's location, calculate the current altitude and horizontal distance between the drone and the target point. Combined with the drone's flight speed, calculate the drop timing during flight. The comparison of the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system further includes: Based on the feature point neighborhood information contained in the feature point descriptor calculated by the SURF algorithm, the K nearest neighbor method is used to obtain the two potential best matching points of each feature point in the target image in the image acquired in real time by the binocular vision system. The best matching point pair is selected by the ratio between the best and second-best distances between the matching points; When the ratio between the best and second-best values ​​is less than the feature intensity threshold, the feature point with the best value is taken as the successfully matched feature point and marked in the image of the binocular vision system.

2. The method for precise throwing with an infrared visual projectile as described in claim 1, characterized in that, The step of calculating the feature points and descriptors of the target point according to the preset algorithm further includes: The SURF algorithm is used to calculate the feature points and feature point descriptors of the target point.

3. The method for precise throwing with an infrared visual projectile as described in claim 2, characterized in that, The calculation of feature points and feature point descriptors for target points based on the SURF algorithm further includes: The target point image is Gaussian smoothed, and a Gaussian scale space is established. The local extrema of the Hessian matrix determinant are calculated to determine the feature point locations of the target point. For any point in the space with scale σ... The Hessian matrix of (x, y) is defined as: In the formula, L xx It is a second-order transverse differential, L xy It is a second-order cross-differentiation, L yy Second-order longitudinal differential; The principal direction of the feature point is calculated, a multi-dimensional feature vector is constructed for the feature point, and the multi-dimensional feature vector is normalized to obtain the feature point descriptor, which contains the neighborhood information of the feature point.

4. The method for precise throwing with an infrared visual projectile as described in claim 1, characterized in that, Calculate the current altitude and horizontal distance between the drone and the target point. Combined with the drone's flight speed, calculate the drop timing during flight, which further includes: Horizontal distance x = V0x * t, where V0x is the horizontal velocity of the drone at the time of deployment, and t is time; Vertical distance y = V0y * t + (1 / 2) * g * t^2, where V0y is the vertical velocity of the drone at the time of deployment, t is time, and g is the acceleration due to gravity. When the vertical velocity V0y of the drone is 0, the vertical distance y = (1 / 2) * g * t^2; The vertical distance between the drone and the target point remains essentially constant, thus determining the time the projected object remains in the air: t = ; The drone's drop point is the horizontal distance from the target point: x = V0x * t = V0x * .

5. The method for precise throwing with an infrared visual projectile as described in claim 1, characterized in that, Calculate the current altitude and horizontal distance between the drone and the target point. Combined with the drone's flight speed, calculate the drop timing during flight, which further includes: Let the air resistance during the fall of the projectile be k. The aerial and horizontal distances of the drone from the target point are as follows: Where x is the horizontal distance, y is the vertical distance, m is the mass of the thrown object, g is the gravitational acceleration, k is the air resistance, V0 is the horizontal velocity of the UAV, and t is the throwing time; By selecting a set of data for m, k, and g, the drop time t is obtained, and then the horizontal distance between the UAV and the target point at the time of drop is determined.

6. A point-to-point throwing system for an infrared visual throwing device, characterized in that, include: An infrared visual projector is mounted on a drone, and the image data collected by its infrared and visible light lenses is connected to the drone's binocular vision system. The drone acquires images including the target point in real time through a forward-facing camera, calculates the feature points of the target point according to a preset algorithm, compares the feature points of the target point with the feature points of the real-time image acquired by the binocular vision system to obtain the location of the target point in the binocular vision system, calculates the current altitude and horizontal distance of the drone from the target point based on the location of the target point, and calculates the release time point during the flight process by combining the drone's flight speed. The drone is further configured as follows: Based on the feature point neighborhood information contained in the feature point descriptor calculated by the SURF algorithm, the K nearest neighbor method is used to obtain the two potential best matching points of each feature point in the target image in the image acquired in real time by the binocular vision system. The best matching point pair is selected by the ratio between the best and second-best distances between the matching points; When the ratio between the best and second-best values ​​is less than the feature intensity threshold, the feature point with the best value is taken as the successfully matched feature point and marked in the image of the binocular vision system.

7. The point-to-point throwing system of the infrared visual throwing device as described in claim 6, characterized in that, The infrared visual projector includes an upper shell module, a mechanism module, and a lower shell module. The mechanism module is fixedly disposed inside the lower shell module, and the upper shell module is fixedly connected to the lower shell module. The mechanism module integrates a heat dissipation unit, a camera unit, a power supply unit, a lens unit, and a padlock unit. The power supply unit is electrically connected to the heat dissipation unit, the camera unit, the lens unit, and the padlock unit. The lens unit includes a lens mounting bracket, an infrared lens, and a visible light lens. The infrared lens and the visible light lens are fixed below the lens mounting bracket. The power supply unit, the camera unit, and the heat dissipation unit are stacked sequentially from bottom to top above the lens mounting bracket. The padlock unit includes a first padlock assembly and a second padlock assembly, the first padlock assembly and the second padlock assembly being disposed opposite to each other and located on both sides of the lens fixing bracket; The first padlock assembly includes an electronic lock and a bracket. The electronic lock is fixedly connected to the power unit via the bracket, and the electronic lock is used to hang a projectile. The bottom of the lower shell module has a slot so that the thrown object can be hooked onto the electronic lock.

8. The fixed-point throwing system of the infrared visual throwing device as described in claim 7, characterized in that, The upper corner of the electronic lock is recessed to form a notch, and a protrusion for triggering the unlocking of the electronic lock is provided on the notch; The lower end of the electronic lock has a groove, and a crossbar that is linked to the protrusion is provided in the groove; In the locked state, the crossbar passes through the groove; when the protrusion is pressed, it causes the crossbar to retract, thus unlocking by pressing.

9. The fixed-point throwing system of the infrared visual throwing device as described in claim 7, characterized in that, The electronic lock is a titanium wire electronic lock, which is controlled by a remote control. When unlocking is required, the titanium wire is heated, causing it to contract instantly and triggering the titanium wire-related structure to unlock, thus enabling aerial throwing.

Images