Mouse disease degree detection system and method based on YOLO

Abstract

The application belongs to the technical field of agricultural intelligence, and provides a mouse disease degree detection system and method based on YOLO, in order to solve the problem of low accuracy of mouse damage detection in farmland in the traditional technology, the method determines the current monitoring image corresponding to the preset target farmland and the corresponding mouse monitoring audio, and determines the preset mouse unique morphological characteristics; in the case that the mouse monitoring audio contains the mouse behavior sound corresponding to the mouse behavior, and according to the preset mouse unique morphological characteristics, in the case of preliminary determination that the current monitoring image contains the mouse image, the current frame mouse unique morphological characteristics corresponding to the current monitoring image is determined; in the case that the current frame mouse unique morphological characteristics conforms to the mouse action morphological trajectory, it is finally determined that the current monitoring image contains the mouse image, and according to a plurality of the above final determinations, the mouse damage degree corresponding to the preset target farmland is determined, which can improve the accuracy of farmland mouse damage detection.

Description

Technical Field

[0001] This invention relates to the fields of agricultural intelligence and machine learning technology, and in particular to a YOLO-based system and method for detecting the severity of rat diseases. Background Technology

[0002] In agriculture, strengthening the detection of rats in farmland and their damage is a crucial factor in ensuring food production security. With the development of intelligent agriculture, artificial intelligence technology is being increasingly applied to the detection of rats and their damage in farmland, aiming to achieve automated and efficient intelligent monitoring of rats and their hazards.

[0003] In intelligent agricultural technology, strengthening the detection of rat damage to farmland requires first identifying rats in the farmland. Traditionally, this is done by collecting images of the farmland and using machine learning-based target detection to detect whether rats are present in the images. When identifying rats, the overall shape and outline of the rat are used as the basis for identification and corresponding detection. Then, based on the detected rats, the degree of damage to the farmland can be identified, thus achieving the detection of the extent of rat damage to the farmland.

[0004] However, the inventors realized that in traditional technologies, because the environment of rats in farmland is complex, their features are not obvious, and their fur color is camouflaged, the target color of the rat is similar to the corresponding background color, which increases the difficulty of detecting rats as targets. In other words, it is difficult to accurately detect rats as targets in farmland, and it is also difficult to accurately detect the harm caused by rats automatically.

[0005] Therefore, improving the accuracy of rodent detection in farmland has become an urgent problem to be solved in the field of intelligent agriculture. Summary of the Invention

[0006] The technical problem solved by this invention is to address the low accuracy of rodent detection in farmland using traditional techniques.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Determine the current monitoring image and corresponding rat monitoring audio corresponding to a preset target farmland, and determine preset rat-specific morphological features; If the rat monitoring audio is determined to contain rat behavior sounds corresponding to rat behavior, and based on the preset rat-specific morphological features and a preset rat target detection YOLO model, if it is initially determined that the current monitoring image contains a rat image, determine the rat-specific morphological features of the current frame corresponding to the current monitoring image; If the rat-specific morphological features of the current frame conform to the rat movement trajectory, finally determine that the current monitoring image contains a rat image, and based on several of the above final determinations, determine the degree of rat infestation corresponding to the preset target farmland.

[0008] Preferably, when it is determined that the unique morphological features of the mouse in the current frame conform to the mouse's movement pattern trajectory, the method includes: determining several adjacent monitoring images based on a time series corresponding to the current monitoring image; determining the unique morphological features of the mouse in adjacent frames corresponding to the adjacent monitoring images, thereby obtaining several unique morphological features of the mouse in adjacent frames; based on the corresponding time series, forming a sequence by combining the unique morphological features of the mouse in the current frame with all the unique morphological features of the mouse in adjacent frames, thereby obtaining a sequence of unique morphological features of the mouse; determining whether the sequence of unique morphological features of the mouse conforms to a preset mouse movement pattern rule; if the above determination is yes, determining that the unique morphological features of the mouse in the current frame conform to the mouse's movement pattern trajectory.

[0009] The present invention also provides a YOLO-based system for detecting the severity of rodent infestation, comprising: a first determining module, configured to determine the current monitoring image and corresponding rodent monitoring audio corresponding to a preset target farmland, and determine preset rodent-specific morphological features; a first determining module, configured to, when determining that the rodent monitoring audio contains rodent behavioral sounds corresponding to rodent behavior, and based on the preset rodent-specific morphological features and a preset rodent target detection YOLO model, determine the current frame rodent-specific morphological features corresponding to the current monitoring image if it is initially determined that the current monitoring image contains a rodent image; and a second determining module, configured to, when determining that the current frame rodent-specific morphological features match the rodent movement trajectory, ultimately determine that the current monitoring image contains a rodent image, and determine the rodent infestation severity corresponding to the preset target farmland based on several of the above final determinations.

[0010] The beneficial effects of this invention are as follows: By determining the current monitoring image and corresponding rat monitoring audio corresponding to the preset target farmland, and determining the preset unique morphological characteristics of the rats, the invention achieves accurate detection of farmland rats based on the unique morphological characteristics of the rats, combined with image recognition-based target detection and voiceprint recognition. This allows for accurate determination of the rat infestation level corresponding to the preset target farmland. It not only realizes a multi-dimensional, cross-modal parallel detection strategy that combines voiceprint recognition with target detection based on unique rat morphological characteristics, but also implements a progressive hierarchical detection strategy of "preliminary judgment - re-judgment verification." Compared to the single-dimensional target detection in traditional technologies, this invention can perform multiple composite verifications of rat detection from different angles, effectively improving the accuracy of farmland rat detection. This adapts to the complex environmental background, indistinct features, and camouflage-like fur of rats in farmland, thereby accurately assessing the harm rats cause to farmland, improving the accuracy of farmland rat infestation detection, and achieving automatic and efficient intelligent monitoring of farmland rats and their harm. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating the YOLO-based method for detecting the severity of rat diseases, as provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the overall process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention.

[0013] Figure 3 This is a schematic diagram of the first sub-process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention.

[0014] Figure 4 This is a schematic diagram of the second sub-process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention.

[0015] Figure 5 This is a schematic block diagram of a YOLO-based system for detecting the severity of rat diseases, provided in an embodiment of the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] This invention provides a YOLO-based system and method for detecting the severity of rodent diseases. The system and method can be applied to devices including but not limited to edge devices, smartphones, tablets, computer equipment, servers, etc., and are used in situations including but not limited to detecting the severity of rodent diseases based on YOLO.

[0018] Example 1, please refer to Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the overall process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention. Figure 1 As shown, in this embodiment, the method includes, but is not limited to, the following steps S101-S103:

[0019] S101. Determine the current monitoring image and corresponding rat monitoring audio corresponding to the preset target farmland, and determine the preset unique morphological characteristics of rats.

[0020] Image and audio acquisition devices are pre-deployed in a predetermined target farmland. The image acquisition devices capture video or images of the target farmland, and based on these images, target detection is used to detect rats and monitor the farmland. Similarly, the audio acquisition devices capture audio from the target farmland and use voiceprint recognition for rat detection, also monitoring rats in the target farmland. The image acquisition devices include, but are not limited to, infrared cameras supporting nighttime shooting, starlight-level low-light cameras, and visible light cameras (RGB cameras). The audio acquisition devices include, but are not limited to, omnidirectional condenser microphones and waterproof, high-sensitivity microphone arrays with directional microphone arrays. Thus, by acquiring video or images and audio corresponding to the target farmland, the current monitoring image and corresponding rat monitoring audio for the target farmland are obtained. Subsequently, rat detection in the target farmland is performed based on the current monitoring image and corresponding rat monitoring audio. The current monitoring image represents the image of the rats in the preset target farmland that is being monitored at the current moment. The corresponding rat monitoring audio represents the audio of the rats in the preset target farmland that is being monitored at the same monitoring spatial target and at the same time as the current monitoring image. For example, if the current monitoring image corresponds to the monitoring spatial target and time B corresponding to location A of the preset target farmland, the corresponding rat monitoring audio represents the audio of the monitoring spatial target and time B corresponding to location A of the preset target farmland.

[0021] Pre-defined morphological features unique to mice are established. These features represent the unique morphological characteristics of mice and can be used to identify mice as distinctive morphological markers. These features include, but are not limited to, unique head, torso, limbs, and tail characteristics, as well as dynamic and static characteristics, and combinations of these features from different angles. Unique head features include, but are not limited to, a pointed snout, small round ears (ear length approximately equal to the distance from the eye to the nose), prominent whiskers, and combinations thereof. Unique torso features include, but are not limited to, a soft, smooth body, a slightly swollen abdomen (more pronounced after eating), a long cylindrical torso, and combinations thereof. Unique limb features include, but are not limited to, short forelimbs with sharp claws (suitable for digging), long, slender hind limbs (extended when jumping), and combinations thereof. Unique tail features include, but are not limited to, a length approximately equal to the body length (house mouse) or longer (e.g.,...). Brown rats (with ring-shaped scales on their surface, thicker base, and tail posture (raised / dragging) reflecting emotional state, such as being raised when alert) and combinations thereof, are among the unique morphological characteristics of rats, including but not limited to the following: back arching during movement, body curled up in a ball when at rest, running posture (e.g., elongated body, alternating extension of hind and hind legs, gait in a "wavy" shape), biting posture (e.g., forepaws holding food, head frequently swaying from side to side), and combinations thereof. The unique morphological characteristics of rats represent the unique structure and state exhibited by the rat. The difference between morphology and shape is that morphology (Form / Morphology) represents the structure and state exhibited by the rat, integrating geometric shape and dynamic changes, while shape (Shape) represents the geometric features of the rat's outer contour, focusing on the static spatial attributes of the rat's outer contour boundary (e.g., length, width, angle, curvature, etc.). Therefore, the meaning and content of the morphological characteristics and shape characteristics of rats are different. This invention's embodiments determine the corresponding unique characteristics from the perspective of rat morphology to perform target detection and target recognition of rats. Furthermore, for several preset mouse-specific morphological features, a preset mouse-specific morphological feature set can be used to represent them. Thus, determining the preset mouse-specific morphological features is expressed as determining the preset mouse-specific morphological feature set, where the preset mouse-specific morphological feature set contains several preset mouse-specific morphological features, thereby more accurately covering several preset mouse-specific morphological features of mice.

[0022] Determine the current monitoring images and corresponding rat monitoring audio corresponding to the preset target farmland, including:

[0023] Identify the rat monitoring videos and corresponding rat monitoring audio for the preset target farmland;

[0024] Determine several single-frame monitoring video images corresponding to the mouse monitoring video;

[0025] Based on a preset image determination method, the current monitoring image corresponding to the preset target farmland is determined from a plurality of single-frame monitoring video images.

[0026] Specifically, a preset image determination method is used, which means determining one frame of monitoring video image from a number of single-frame monitoring video images corresponding to the mouse monitoring video as the current monitoring image corresponding to the preset target farmland. The preset image determination method can be a random method, that is, randomly determining one frame of monitoring video image from a number of single-frame monitoring video images corresponding to the mouse monitoring video as the current monitoring image corresponding to the preset target farmland. Alternatively, the preset image determination method can be to take the first frame of monitoring video image from a number of single-frame monitoring video images corresponding to the mouse monitoring video as the current monitoring image corresponding to the preset target farmland. The preset image determination method is not limited here.

[0027] Based on the above settings, the rat monitoring video and its corresponding rat monitoring audio corresponding to the preset target farmland are determined. The rat monitoring video and its corresponding rat monitoring audio constitute the audio and video of the preset target farmland, i.e., audio and video of the same target in the same monitoring space and within the same time period. Then, the rat monitoring video is converted into corresponding single-frame images, thereby determining several single-frame monitoring video images corresponding to the rat monitoring video. A single-frame monitoring video image represents a single frame of image based on the monitoring video. Several single-frame monitoring video images are generally time-series single-frame images that correspond to the time sequence of the rat monitoring audio. Then, based on the preset image determination method, the current monitoring image corresponding to the preset target farmland is determined from the several single-frame monitoring video images. Thus, by determining the current monitoring image corresponding to the preset target farmland based on the video stream of the rat monitoring video and its corresponding rat monitoring audio, the temporal continuity of the several single-frame monitoring video images and their correspondence with the rat monitoring audio can be ensured, so that subsequent rat detection can be performed based on the corresponding several single-frame monitoring video images and rat monitoring audio, thereby improving the effectiveness and accuracy of rat detection.

[0028] S102. If it is determined that the mouse monitoring audio contains the sound of mouse behavior corresponding to the mouse behavior, and based on the preset mouse-specific morphological features and the preset mouse target detection YOLO model, if it is initially determined that the current monitoring image contains a mouse image, determine the mouse-specific morphological features of the current frame corresponding to the current monitoring image.

[0029] A pre-set YOLO model for mouse target detection is defined as a YOLO model based on mice as the detection target. The YOLO model, or YOLO (You Only Look Once) model, is a real-time target detection model. The pre-set YOLO model for mouse target detection includes, but is not limited to, YOLOv3 to YOLOv8.

[0030] A pre-set mouse behavior voiceprint recognition model is used to identify the sounds corresponding to mouse behaviors. The pre-set mouse behavior voiceprint recognition model includes, but is not limited to, deep neural networks (DNN), convolutional neural networks (CNN), and recurrent neural networks (RNN). The sounds corresponding to mouse behaviors include, but are not limited to, the squeaking sounds of mice interacting with each other, the screaming sounds of mice being attacked, the gnawing sounds of mice biting, and the rubbing sounds of mice moving.

[0031] Based on the above description, mice are identified by using preset mouse-specific morphological features as the detection target. Based on the preset mouse target detection YOLO model, it is initially determined whether the current monitoring image contains a mouse image, that is, whether the current monitoring image contains the corresponding preset mouse-specific morphological features. Based on the preset mouse behavior voiceprint recognition model, it is determined whether the mouse monitoring audio contains the mouse behavior sound corresponding to the mouse behavior. The mouse behavior sound represents the sound corresponding to the mouse behavior, as described above.

[0032] If the audio of the mouse monitoring system is determined to contain the sounds of mouse behavior, and the current monitoring image is initially determined to contain a mouse image, since mouse behavior is generally accompanied by sound and morphology, this indicates that the current monitoring image is highly likely to contain a mouse image, but further verification is needed. Therefore, the unique morphological features of the mouse in the current frame corresponding to the current monitoring image are determined. The unique morphological features of the mouse in the current frame represent the unique morphological features of the mouse contained in the current monitoring image. The unique morphological features of the mouse in the current frame can be a specific preset unique morphological feature of the mouse mentioned above, or it can be a set of several preset unique morphological features of the mouse. There is no limitation here. The unique morphological features of the mouse in the current frame corresponding to the current monitoring image can be obtained by outputting the corresponding features when performing target detection based on the preset mouse target detection YOLO model.

[0033] Similarly, if it is initially determined that the current single-frame monitoring video image does not contain a mouse image, or the mouse monitoring audio does not contain the sound of mouse behavior, it indicates that the current monitoring image is highly likely not to contain a mouse image. It is assumed by default that the current monitoring image does not contain a mouse image, and the unique morphological features of the current frame corresponding to the current single-frame monitoring video image are uncertain, that is, no further verification is required.

[0034] In summary, by combining voiceprint recognition with target detection based on mouse-specific morphological features, a multi-dimensional, cross-modal parallel detection strategy can improve the accuracy of the preliminary judgment and thus effectively improve the accuracy of farmland mouse detection.

[0035] S103. If the unique morphological features of the rat in the current frame match the rat's movement pattern trajectory, it is finally determined that the current monitoring image contains a rat image, and based on several of the above final determinations, the degree of rat infestation corresponding to the preset target farmland is determined.

[0036] Since rat behavior is generally a continuous expression of sound and form over a certain period of time, after initially determining that the current monitoring image contains a rat image, we then judge from the perspective of the continuous morphological expression pattern corresponding to the above continuous expression (i.e., the rat's movement morphological trajectory) whether the rat's unique morphological features in the current frame of the current monitoring image match the rat's movement morphological trajectory, so as to verify whether the current monitoring image does indeed contain a rat image.

[0037] Based on the above description, it is determined whether the unique morphological features of the mouse in the current frame conform to the mouse's motion morphological trajectory. The mouse motion morphological trajectory represents a set of temporal or sequential expressions of the mouse's corresponding actions in terms of morphology. For example, in the case of a mouse running continuously, since the mouse's running posture is characterized by an elongated body and alternating extension of the hind and forelegs (a "wavy" gait), the mouse's running motion morphological trajectory includes at least temporal alternating extension of the hind and forelegs (a "wavy" gait). In other words, the temporal expression of alternating hind and foreleg extension (a "wavy" gait) constitutes the mouse's running motion morphological trajectory. The unique morphological features of the mouse in a single frame of a monitoring image can be: the unique morphological feature of the mouse in one monitoring image is foreleg extension; the unique morphological feature of the mouse in the next monitoring image is hind leg extension; the unique morphological feature of the mouse in the next monitoring image is again foreleg extension, and so on. The alternation of foreleg extension and hind leg extension is the temporal expression of the mouse's morphological features and patterns corresponding to continuous running. Other examples follow the same pattern.

[0038] Therefore, if the unique morphological features of the mouse in the current frame are determined to match the mouse's movement trajectory, it indicates that, from the perspective of the continuous morphological expression rules corresponding to the aforementioned continuous expression (i.e., the mouse's movement trajectory), the occurrence of the unique morphological features of the mouse in the current frame corresponding to the current monitoring image is consistent with the continuous morphological expression rules corresponding to the mouse's corresponding behavior. That is, from the perspective of the continuous morphological expression rules corresponding to the aforementioned continuous expression (i.e., the mouse's movement trajectory), the current monitoring image should correspond to the unique morphological features of the mouse in the current frame. For example, in the case of a mouse running continuously, since the mouse's running posture is characterized by an elongated body and... The alternating extension of the forelegs and hind legs (a "wavy" gait) indicates that the running motion trajectory of a mouse includes at least this alternating extension of the forelegs and hind legs (a "wavy" gait). Therefore, assuming that the unique morphological feature of the mouse in the previous frame corresponding to the running posture in the previous monitoring image is foreleg extension, according to the running posture pattern, the unique morphological feature of the mouse in the current frame corresponding to the running posture in the current monitoring image should be hind leg extension. If this is indeed the case, meaning the unique morphological feature of the mouse in the current frame is indeed hind leg extension, it indicates that the unique morphological feature of the mouse in the current frame matches the mouse's motion trajectory, further proving the preliminary judgment that the current monitoring image contains... The image of a mouse is more likely to be the correct judgment, thus defaulting to the final judgment that the current monitored image contains a mouse image. Based on several of the above final judgments, the degree of rodent infestation corresponding to the preset target farmland is determined. The degree of rodent infestation indicates the extent of harm caused by mice to the preset target farmland, that is, the degree of disease caused by mice. The degree of rodent infestation can be measured by the size of the mouse population (i.e., the number of mice detected in the preset target farmland), the activity level of the mice, or a combination of the above indicators. Otherwise, it indicates that the current monitored image is determined to contain a mouse image from the perspective of the continuous morphological expression pattern corresponding to the above continuous expression (i.e., the morphological trajectory of mouse movement). The occurrence of the unique morphological features of the mouse in the current frame does not conform to the continuous morphological expression pattern corresponding to the corresponding behavior of the mouse. That is, from the perspective of the continuous morphological expression pattern (i.e., the morphological trajectory of mouse movement) corresponding to the above continuous expression, the current monitoring image should not correspond to the unique morphological features of the mouse in the current frame. This indicates that the above preliminary judgment may have been misjudged. Finally, it is determined that the current monitoring image does not contain a mouse image, thus correcting the preliminary judgment. In this way, a progressive hierarchical detection strategy of "preliminary judgment - re-judgment verification" is realized, which can perform composite verification of mouse detection from different angles and effectively improve the accuracy of farmland mouse detection.

[0039] In this embodiment of the invention, the current monitoring image and corresponding rat monitoring audio corresponding to a preset target farmland are determined, and preset rat-specific morphological features are determined. If the rat monitoring audio is determined to contain rat behavior sounds corresponding to rat behavior, and based on the preset rat-specific morphological features and a preset rat target detection YOLO model, if it is initially determined that the current monitoring image contains a rat image, then the current frame rat-specific morphological features corresponding to the current monitoring image are determined. If the current frame rat-specific morphological features match the rat movement trajectory, then the current monitoring image is finally determined to contain a rat image. Based on several of the above final determinations, the target farmland is determined. This technology assesses the severity of rodent damage, enabling a multi-dimensional, cross-modal parallel detection strategy that combines voiceprint recognition with target detection based on unique rat morphological features. Furthermore, it implements a progressive hierarchical detection strategy of "preliminary judgment - further judgment verification." Compared to single-dimensional target detection in traditional technologies, this approach allows for composite verification of rat detection from different angles, effectively improving the accuracy of farmland rat detection. It adapts to the complex environmental backgrounds, indistinct features, and camouflage-like fur of rats in farmland, thereby accurately assessing the harm rats cause to farmland, improving the accuracy of farmland rodent damage detection, and achieving automatic and efficient intelligent monitoring of farmland rats and their damage.

[0040] In one embodiment, please refer to Figure 3 , Figure 3 This is a schematic diagram of the first sub-process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention. Figure 3 As shown, in this embodiment, when it is determined that the unique morphological features of the mouse in the current frame match the mouse's movement morphological trajectory, the following steps are taken:

[0041] S301. Determine several adjacent monitoring images based on time series corresponding to the current monitoring image;

[0042] S302. Determine the unique morphological features of the mouse in the adjacent frames corresponding to the monitored adjacent images, and obtain several unique morphological features of the mouse in the adjacent frames.

[0043] S303. Based on the corresponding time series, the mouse-specific morphological features of the current frame and the mouse-specific morphological features of all adjacent frames are combined to form a sequence, thereby obtaining a mouse-specific morphological feature sequence.

[0044] S304. Determine whether the unique morphological feature sequence of the mouse conforms to the preset mouse action pattern corresponding to the mouse.

[0045] S305. If the above judgment is correct, it is determined that the unique morphological features of the mouse in the current frame conform to the mouse action morphological trajectory.

[0046] S306. If the above judgment is not true, it is determined that the unique morphological features of the mouse in the current frame do not conform to the mouse action morphological trajectory.

[0047] The pre-set pattern of mouse movement patterns, or preset pattern of mouse movement patterns, means that the corresponding movement of the mouse is expressed in a morphological manner based on time sequence or order. For example, as mentioned above, when the mouse runs continuously, the alternating extension of the front and hind legs corresponds to the alternating extension of the front and hind legs. The alternating appearance of the unique morphological features of the mouse's front and hind legs in several monitoring images based on time sequence is the content of the preset pattern of the front and hind legs of the mouse running pattern.

[0048] To determine whether the unique morphological features of the mouse in the current frame conform to the mouse's movement trajectory, i.e., to determine whether the unique morphological features of the mouse in the current frame appear at the appropriate time, a sequence of unique morphological features of the mouse corresponding to several adjacent frames of monitoring images is constructed from a time-series-based sequential expression perspective. Then, it is determined whether the sequence of unique morphological features of the mouse conforms to the morphological manifestation rules of the corresponding actions based on a time-series or sequential set expression. Thus, the relationship judgment between the unique morphological features of the mouse in the current frame (i.e., the "point" dimension) and the mouse's movement trajectory (i.e., the "set" dimension) is transformed into a set relationship judgment between the unique morphological feature sequence of the mouse (i.e., the "set" dimension) and the preset mouse movement pattern (i.e., the "set" dimension). This not only transforms the judgment dimension to facilitate judgment, but also performs a continuous judgment corresponding to the duration of the action based on several continuous monitoring images. By using the continuous judgment, the judgment of the action logic is expanded from the dimension of the continuous process before and after, which can improve the accuracy of the corresponding judgment.

[0049] Based on the above setup and concept, several time-series-based adjacent monitoring images are determined for the current monitoring image. These adjacent monitoring images represent monitoring images that are adjacent to the current monitoring image. These adjacent monitoring images can be monitoring images preceding the current monitoring image (preceding it), following it (following it), or in the middle of it (being preceding it). The process is then repeated to determine the current frame corresponding to the current monitoring image. By identifying mouse-specific morphological features that are identical or similar, the mouse-specific morphological features corresponding to adjacent images in each monitoring frame are determined, resulting in several mouse-specific morphological features for adjacent frames. These adjacent frame mouse-specific morphological features represent the mouse-specific morphological features corresponding to each monitoring image in each monitoring frame. Then, based on the corresponding time series, that is, based on the time order of the current monitoring image and all monitoring adjacent images, the mouse-specific morphological features of the current frame are combined with the mouse-specific morphological features of all adjacent frames to form a sequence, resulting in a mouse-specific morphological feature sequence. This mouse-specific morphological feature sequence represents a set of regular expressions of several mouse-specific morphological features based on the time series. The determination of whether a mouse-specific morphological feature sequence conforms to a pre-defined mouse movement pattern is crucial. For example, if the mouse-specific morphological feature sequence represents a running posture, the pre-defined mouse movement pattern is the set of regular expressions corresponding to the alternating extension of the forelegs and hindlegs. Therefore, the determination of whether the mouse-specific morphological feature sequence conforms to the regular expression corresponding to the alternating extension of the forelegs and hindlegs is crucial. In other words, the determination of whether the mouse-specific morphological feature sequence is expressed as follows: one frame of the monitoring image shows foreleg extension, the next frame shows hind leg extension, and so on. The corresponding mouse-specific morphological features are foreleg extension, ..., and the mouse-specific morphological features corresponding to foreleg extension and hind leg extension appear alternately. If the above judgment is true, it is determined that the mouse-specific morphological features of the current frame conform to the mouse's movement trajectory. If the above judgment is false, it is determined that the mouse-specific morphological features of the current frame do not conform to the mouse's movement trajectory. This not only transforms the judgment dimension to facilitate judgment, but also performs a continuous judgment corresponding to the duration of the action based on several continuous monitoring images. By using the continuous judgment, the judgment of the action logic is expanded from the dimension of the continuous process before and after, which can improve the accuracy of the corresponding judgment.

[0050] In this embodiment of the invention, the unique morphological feature sequence is judged to determine whether it conforms to the preset mouse movement pattern. This is done by judging whether the unique morphological features of the mouse in the current frame conform to the mouse movement pattern trajectory. From the perspective of the mouse movement pattern, and with the help of the unique morphological features of the mouse in adjacent frames based on the time series, the current monitoring image is re-verified to see if it contains a mouse image. This verifies whether the initial judgment that the current monitoring image contains a mouse image is a misjudgment. Since the mouse must also appear in the adjacent monitoring images in the adjacent frames if the current monitoring image and the adjacent monitoring images in a single frame each contain a mouse, it is also impossible to make a misjudgment regarding whether the unique morphological feature sequence based on the time series conforms to the preset mouse movement pattern. Therefore, by using the composite judgment and detection strategy of "preliminary judgment - re-judgment verification", the accuracy of farmland mouse detection can be effectively improved compared with the single-dimensional target detection in traditional technology. This is to adapt to the environmental characteristics of mice in farmland, such as complex backgrounds, indistinct features, and camouflage fur.

[0051] In one embodiment, when it is determined that the mouse monitoring audio contains mouse behavior sounds corresponding to mouse behavior, and based on preset mouse-specific morphological features and a preset mouse target detection YOLO model, if it is initially determined that the current monitoring image contains a mouse image, the mouse-specific morphological features of the current frame corresponding to the current monitoring image are determined, including:

[0052] Based on the preset unique morphological characteristics of mice and the current monitoring image, and using the preset mouse target detection YOLO model, the system detects whether the current monitoring image contains a mouse image.

[0053] Based on a pre-defined rat behavior voiceprint recognition model, it identifies whether the rat monitoring audio contains the rat behavior sounds corresponding to the rat behavior.

[0054] If the current monitoring image contains a mouse image and the mouse monitoring audio contains mouse behavioral sounds, determine the unique morphological features of the mouse in the current frame corresponding to the current monitoring image.

[0055] If no mouse image is detected in the current monitoring image, or no mouse behavior sound is identified in the mouse monitoring audio, the unique morphological features of the mouse in the current frame corresponding to the current monitoring image are uncertain.

[0056] As described above, a YOLO model for mouse target detection and a pre-set mouse behavior voiceprint recognition model are configured. Based on the pre-set mouse target detection YOLO model, it is initially determined whether the current monitoring image contains a mouse image. Based on the pre-set mouse behavior voiceprint recognition model, it is determined whether the mouse monitoring audio contains the sound corresponding to the mouse behavior. If the current monitoring image contains a mouse image and the mouse monitoring audio contains the sound of the mouse behavior (i.e., the mouse is identified from both target detection and sound recognition), the unique morphological features of the mouse in the current frame corresponding to the current monitoring image are determined to further verify the target detection. If the current monitoring image does not contain a mouse image or the mouse monitoring audio does not contain the sound of the mouse behavior (i.e., the mouse is not identified from both target detection and sound recognition), the unique morphological features of the mouse in the current frame corresponding to the current monitoring image are not determined, and no further verification of the target detection is performed.

[0057] This invention implements a detection strategy that combines voiceprint recognition with target detection based on mouse-specific morphological features to determine whether further verification is needed. Since the morphology and sound of a mouse's behavior are generally closely linked, i.e., sound and form are interdependent, combining voiceprint recognition with image-based target detection allows for a multi-dimensional, cross-modal parallel detection strategy that combines voiceprint recognition with target detection based on mouse-specific morphological features when detecting whether the current monitoring image contains a mouse image. This improves the accuracy of the initial judgment and allows for further verification. By performing multiple composite verifications of mouse detection from different perspectives, the overall accuracy of farmland mouse detection is effectively improved.

[0058] In one embodiment, based on preset mouse-specific morphological features and the current monitoring image, and using a preset mouse target detection YOLO model, the detection of whether the current monitoring image contains a mouse image includes:

[0059] Determine a preset set of unique morphological features of mice, wherein the preset set of unique morphological features of mice includes at least one of the following: a preset set of unique morphological features of mouse head, a preset set of unique morphological features of mouse trunk, and a preset set of unique morphological features of mouse tail.

[0060] Based on the preset YOLO model for mouse target detection, the system detects whether the current monitored image contains preset unique morphological features of mouse head, mouse trunk, or mouse tail.

[0061] If the above detection is positive, it is determined that the currently monitored image contains an image of a mouse.

[0062] If the above detection is negative, it is determined that no image of a mouse was detected in the currently monitored image.

[0063] Based on the morphological characteristics of mice, their unique morphological features are divided into three parts: head, torso, and tail. This involves presupposing unique morphological features for the mouse head, torso, and tail, forming a set of presupposed unique mouse morphological features. This set includes at least one of the following: unique morphological features for the mouse head, torso, or tail. Specifically, the unique morphological features for the mouse head include, but are not limited to, a pointed snout, small round ears (ear length approximately equal to the distance from the eye to the nose), prominent whiskers, and a biting posture (forehand paws holding food, head...). The mouse's body features include, but are not limited to, a soft, smooth body without sharp edges, a slightly swollen abdomen, short forelimbs with sharp claws, slender hind limbs, and a running posture (elongated body with alternating extension of fore and hind legs (a "wavy" gait)). The mouse's tail features include, but are not limited to, a length approximately equal to or longer than the body length (house mouse) or longer (brown mouse), ring-shaped scales on the surface, a thicker base, and a tail posture (raised / dragging on the ground). Based on the single mouse-specific morphological feature or the combination of several mouse-specific morphological features included in the preset set of mouse-specific morphological features, the mouse can be identified.

[0064] Based on the above description, a preset set of unique mouse morphological features is determined, wherein the preset set of unique mouse morphological features includes at least one of the following: a preset unique mouse head morphological feature, a preset unique mouse torso morphological feature, and a preset unique mouse tail morphological feature; and based on a preset mouse target detection YOLO model, it is detected whether the current monitoring image contains a single morphological feature or a combination of morphological features that can identify a mouse, corresponding to the preset unique mouse head morphological feature, the preset unique mouse torso morphological feature, or the preset unique mouse tail morphological feature. If the above detection is yes, it is determined that the current monitoring image contains a mouse image; similarly, if the above detection is no, it is determined that the current monitoring image does not contain a mouse image.

[0065] To achieve more accurate target detection through cross-verification during the target detection process, a YOLO model for mouse target detection is used. When detecting whether the current monitored image contains unique morphological features of the mouse head, torso, or tail, the model further checks whether the image contains unique morphological features of at least two of these features. This comprehensive verification detection involves repeated detection from unique morphological features of different mouse parts. In other words, if a mouse is detected using unique morphological features from one part, and simultaneously detected using unique morphological features from another part, the repeated superposition of both methods results in successful detection of the mouse target. This further improves the accuracy of mouse target detection. However, compared to traditional techniques that detect the entire mouse outline, mouse target detection remains relatively simple.

[0066] Therefore, when performing mouse target detection, the traditional method of identifying mice by detecting their overall outline is transformed into detecting unique local morphological features. This method uses points instead of surfaces and parts instead of the whole to identify and detect mice, eliminating the need to judge all overall features. This improves the flexibility and accuracy of mouse detection, making it more suitable for the complex backgrounds, indistinct features, and camouflage characteristics of mice in farmland environments.

[0067] It should be noted that the difference between preset mouse-specific morphological features and key point features is that, in the embodiments of the present invention, a single preset mouse-specific morphological feature can perform target recognition and detection of the entire mouse, while key point features in traditional technology only represent key points of the mouse in a local area. A single key point feature cannot represent the entire mouse and cannot be used to perform target recognition and detection of the entire mouse. A single key point feature must be combined with other key point features to perform target recognition and detection of the entire mouse.

[0068] In this embodiment of the invention, the detection of whether the current monitoring image contains a mouse image is transformed into "detecting whether the current monitoring image contains preset mouse head-specific morphological features, preset mouse torso-specific morphological features, or preset mouse tail-specific morphological features." Since the preset mouse-specific morphological features represent a high degree of recognizability by identifying mice based on their unique morphological characteristics, that is, seeing the corresponding preset mouse-specific morphological features is enough to identify a mouse. Thus, it realizes mouse detection by replacing the whole mouse with a local unique morphological feature that is mouse-identifiable, and by using a point-to-area approach to mouse detection. Compared with the target detection of the whole shape of the mouse in traditional technology, it can improve the flexibility and robustness of mouse detection in farmland, and adapt to the environmental characteristics of mice in farmland, such as complex backgrounds, indistinct features, and camouflage fur coloration. It can effectively improve the accuracy of mouse detection in farmland.

[0069] In one embodiment, based on a preset mouse behavior voiceprint recognition model, identifying whether mouse monitoring audio contains mouse behavior sounds corresponding to mouse behavior includes at least one of the following:

[0070] The detection function checks whether the audio from the mouse monitoring system contains squeaking sounds corresponding to the mouse's interactive behavior.

[0071] The detection system checks whether the audio from the mouse monitoring system contains the squealing sounds of a mouse that correspond to a perceived act of aggression.

[0072] The detection function checks whether the audio from the mouse monitoring system contains the sounds of mice gnawing.

[0073] The detection function checks whether the audio from the mouse monitoring system contains the rubbing sounds corresponding to the mouse's actions.

[0074] The detection function checks whether the audio recordings of the rat monitoring system contain sounds corresponding to the rat's diseased behaviors.

[0075] Based on a pre-defined rat behavior voiceprint recognition model, identify whether the rat monitoring audio contains rat behavior sounds corresponding to the rat behavior, including at least one of the following:

[0076] The detection method checks whether the audio of the mouse monitoring contains mouse squeaking sounds corresponding to mouse interactive behaviors. These mouse interactive behaviors include, but are not limited to, behaviors such as young mice begging for food and social interactions, which correspond to squeaking sounds. The characteristics of mouse squeaking sounds include, but are not limited to, short, high-frequency, single, or repetitive sounds.

[0077] The detection system checks whether the audio of the mouse monitoring contains the squealing sound corresponding to the mouse being attacked. The mouse being attacked includes, but is not limited to, pain and predator attacks. The squealing sound is characterized by, but is not limited to, being sharp, high-intensity, and lasting 0.5 to 2 seconds.

[0078] The detection system checks whether the audio from the rat monitoring system contains rat gnawing sounds corresponding to rat gnawing behavior. Rats gnawing behavior includes, but is not limited to, gnawing on crops or building materials. Corresponding gnawing sounds are characterized by, but are not limited to, intermittent, wideband, and rhythmic patterns.

[0079] The detection method checks whether the audio of the mouse monitoring contains the friction sounds corresponding to the mouse's actions. The mouse's actions include, but are not limited to, running, digging, or crossing vegetation, which correspond to friction sounds. The characteristics of the mouse friction sounds include, but are not limited to, low-frequency rustling sounds and irregularity.

[0080] The detection system checks whether the audio of the mouse monitoring contains sounds corresponding to mouse disease behaviors. These behaviors include, but are not limited to, respiratory infections, which correspond to high-frequency screams or coughs. The sounds of mouse disease include, but are not limited to, coughing / sneezing.

[0081] Since the physical characteristics and sounds associated with a mouse's behavior are generally closely linked, combining mouse target detection with corresponding audio recognition can further improve the accuracy of mouse target detection.

[0082] In this embodiment of the invention, by utilizing sound detection covering different living scenarios of mice, since the form corresponding to the behavior of mice is generally closely linked to the sound, that is, the form produced by the corresponding behavior of mice must correspond to the corresponding sound, i.e., sound and form are interdependent, the voiceprint recognition is combined with the target detection based on image recognition. When detecting whether the current monitoring image contains a mouse image, a multi-dimensional cross-modal parallel detection strategy that combines voiceprint recognition with target detection based on the unique morphological features of mice can be implemented. This can improve the accuracy of the above preliminary judgment, and then a second judgment can be made to verify it. Multiple composite verifications of mouse detection are performed from different angles, thereby effectively improving the overall accuracy of farmland mouse detection.

[0083] In one embodiment, determining the current monitoring image corresponding to a preset target farmland includes:

[0084] Determine several types of current initial monitoring images for different spectral bands corresponding to the preset target farmland. The several types of current initial monitoring images include at least two of the following: visible light current initial monitoring image, near-infrared current initial monitoring image, and thermal infrared current initial monitoring image;

[0085] Align and stitch together all the current initial monitoring images to obtain the current monitoring image corresponding to the preset target farmland.

[0086] Several types of current initial monitoring images in different spectral bands corresponding to the preset target farmland are determined. These several types of current initial monitoring images include at least two of the following: visible light current initial monitoring image, near-infrared current initial monitoring image, and thermal infrared current initial monitoring image. The visible light current initial monitoring image refers to the current initial monitoring image based on visible light (RGB), which can be acquired based on RGB image or RGB video acquisition equipment. The near-infrared current initial monitoring image refers to the current initial monitoring image based on near-infrared light (NIR), which can be acquired based on near-infrared camera. The thermal infrared current initial monitoring image refers to the current initial monitoring image based on thermal infrared light (TIR), which can be acquired based on thermal infrared (TIR) ​​camera. Thus, several types of current initial monitoring images in different spectral bands corresponding to the preset target farmland are determined. All current initial monitoring images are aligned and stitched together to obtain the current monitoring image corresponding to the preset target farmland.

[0087] Aligning and stitching all the initial monitoring images involves geometrically correcting and spatially matching images from different spectral bands (such as RGB, near-infrared, and thermal infrared) to ensure precise pixel-level overlap and generate a fused image representation. Image alignment uses geometric transformations (such as translation, rotation, and scaling) to ensure that the same scene point in different images completely overlaps in pixel coordinates, eliminating offsets caused by differences in sensor position, viewing angle, or time. For example, aligning the mouse hotspot in a thermal infrared image with the mouse outline in a visible light image. Image stitching merges the aligned multi-band images into a single multi-channel image (such as an RGB + near-infrared 4-channel image) to generate a fused image representation, preserving information from each band for subsequent model processing.

[0088] In this embodiment of the invention, by aligning and stitching together several current initial monitoring images from different spectral bands, multispectral fusion image enhancement is achieved. Rodent pest monitoring can be upgraded from "single target detection" to an image enhancement strategy of "target-environment collaborative analysis," which can improve the robustness of rat target detection. This adapts to the environmental characteristics of rats in farmland, such as complex backgrounds, indistinct features, and camouflage fur. It can further improve the accuracy of farmland rat detection, thereby accurately assessing the damage rats cause to farmland, significantly improving the precision and efficiency of agricultural management, and thus achieving automatic and efficient intelligent monitoring of farmland rats and their damage.

[0089] In one embodiment, please refer to Figure 4 , Figure 4 This is a schematic diagram of the second sub-process of the YOLO-based method for detecting the severity of rat diseases provided in an embodiment of the present invention. Figure 4 As shown, in this embodiment, several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland are determined, including:

[0090] S401. Respond to different preset acquisition trigger events, and based on the corresponding preset video acquisition equipment, acquire several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland.

[0091] S402. Determine a number of single-frame monitoring video images corresponding to each type of mouse monitoring video;

[0092] S403. Based on the preset image determination method and the same time sequence, determine the current initial monitoring image of each type corresponding to the preset target farmland from several single-frame monitoring video images of each type, and obtain several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland.

[0093] Pre-set acquisition trigger events, also known as preset acquisition trigger events, refer to events that trigger the acquisition of rat monitoring videos corresponding to preset target farmland. Pre-set acquisition trigger events include, but are not limited to, time-triggered events, event-triggered events, and combinations of the above-mentioned factors. Time-triggered events refer to the acquisition being triggered when the peak rat activity times are reached, including but not limited to, 2 hours after dusk (18:00-20:00 after sunset), 1-2 hours before dawn (03:00-05:00), and the new moon. Event-triggered events refer to the acquisition being triggered when the corresponding events occur, including but not limited to, cloudy days, underground sensors detecting digging signals, and traps (such as pedal-type rat traps) placed on "rat trails" detecting vibration signals.

[0094] Furthermore, for different preset acquisition trigger events, corresponding preset video acquisition devices are deployed. For example, for preset acquisition trigger events corresponding to sunny daytime, visible light video acquisition devices (such as RGB cameras) and near-infrared video acquisition devices are deployed; for preset acquisition trigger events corresponding to sunny nighttime without moonlight, thermal infrared video acquisition devices and preset short-wave infrared video acquisition devices are deployed, and so on. This allows for the deployment of corresponding video acquisition devices for different acquisition environments and conditions, ensuring that the acquisition of rat monitoring videos is adapted to the corresponding acquisition conditions. This is to adapt to the environmental characteristics of rats in farmland, where the background is complex, features are not obvious, and fur coloration is camouflaged. This can further improve the accuracy of rat detection in farmland and thus accurately assess the harm rats cause to farmland.

[0095] As described above, in response to different preset acquisition trigger events, the corresponding preset video acquisition device is activated, and based on the corresponding preset video acquisition device, several types of rat monitoring videos of different spectral bands corresponding to the preset target farmland are acquired. Several types of rat monitoring videos of different spectral bands can be acquired based on the same video acquisition device that supports the acquisition of several types of rat monitoring videos of different spectral bands, or they can be acquired based on different video acquisition devices for rat monitoring videos of corresponding spectral band types. For example, visible light RGB rat monitoring videos and near-infrared (NIR) rat monitoring videos can be acquired based on a dual-spectrum camera, or an RGB camera can be used to acquire visible light RGB rat monitoring videos and a near-infrared camera can be used to acquire near-infrared (NIR) rat monitoring videos, and so on.

[0096] Based on several types of rat monitoring videos corresponding to different spectral bands of a preset target farmland, for each type of rat monitoring video corresponding to each spectral band, several single-frame monitoring video images are determined for each type of rat monitoring video, that is, the video is converted into several single-frame monitoring video images.

[0097] Based on the preset image determination method and the same time sequence, the current initial monitoring image corresponding to each type of preset target farmland is determined from several single-frame monitoring video images of each type, thereby obtaining several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland. In other words, according to the same image determination method, several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland are determined from several single-frame monitoring video images of each type.

[0098] Responding to different preset acquisition trigger events, and based on corresponding preset video acquisition devices, acquiring several types of rat monitoring videos in different spectral bands corresponding to preset target farmland, including at least one of the following:

[0099] When a preset acquisition trigger event corresponds to preset good lighting conditions, based on preset visible light video acquisition equipment and preset near-infrared video acquisition equipment, several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland are acquired.

[0100] When the preset acquisition trigger event corresponds to the preset low light conditions, based on the preset RGB-NIR multispectral camera video acquisition device and the preset thermal infrared video acquisition device, several types of mouse monitoring videos corresponding to the preset target farmland in different spectral bands are acquired.

[0101] Under preset conditions of no light, the preset acquisition trigger event corresponds to preset conditions of no light. Based on preset thermal infrared video acquisition equipment and preset shortwave infrared video acquisition equipment, several types of rat monitoring videos of different spectral bands corresponding to the preset target farmland are acquired.

[0102] Specifically, in response to different preset acquisition trigger events, and based on corresponding preset video acquisition devices, several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland are acquired, including at least one of the following:

[0103] When the preset acquisition trigger event corresponds to preset good lighting conditions, such as a sunny day, several types of mouse monitoring videos of different spectral bands corresponding to the preset target farmland are acquired using preset visible light video acquisition devices (e.g., RGB cameras) and preset near-infrared video acquisition devices (e.g., NIR cameras). This allows for the acquisition of visible light images to detect mouse morphology, and the capture of near-infrared bands (high reflectivity in healthy vegetation, low reflectivity in rodent-infested areas) to utilize the sunlight reflection spectrum to detect mouse activity traces and crop damage, while also enhancing the contrast between the mice and the crop background. Similarly, when the acquisition trigger event corresponds to preset low light conditions, such as... When the preset acquisition trigger event corresponds to dusk or cloudy weather, based on preset RGB-NIR multispectral camera video acquisition equipment and preset thermal infrared video acquisition equipment, several types of mouse monitoring videos corresponding to different spectral bands of the preset target farmland are acquired under video acquisition conditions corresponding to dusk or cloudy weather. When the preset acquisition trigger event corresponds to preset no-light conditions, for example, when the preset acquisition trigger event corresponds to nighttime no-light conditions, based on preset thermal infrared video acquisition equipment and preset shortwave infrared video acquisition equipment, several types of mouse monitoring videos corresponding to different spectral bands of the preset target farmland are acquired under video acquisition conditions corresponding to nighttime no-light conditions.

[0104] This invention, through event-driven mechanisms, employs corresponding preset video acquisition devices to acquire several types of rat monitoring videos in different spectral bands corresponding to a preset target farmland for different preset acquisition trigger events. This allows for the determination of several types of current initial monitoring images in different spectral bands corresponding to the preset target farmland. The invention can activate appropriate preset video acquisition devices based on different acquisition trigger event conditions, thereby adapting the acquisition of rat monitoring videos to the corresponding acquisition conditions. This improves the acquisition quality of rat monitoring videos, resulting in better rat detection performance and further enhancing the accuracy of rat detection in farmland.

[0105] In one embodiment, determining the degree of rodent infestation corresponding to a preset target farmland based on several of the aforementioned final determinations includes:

[0106] Based on the aforementioned final determinations, statistical indicators of rat activity intensity corresponding to the preset target farmland are calculated, including at least one of the following:

[0107] Determine whether the rat activity intensity index is greater than or equal to the preset first rat activity intensity index threshold, and if the above determination is yes, determine that the rat infestation level corresponding to the preset target farmland is in a high-risk state.

[0108] Determine whether the rat activity intensity index is less than the preset second rat activity intensity index threshold. If the above determination is yes, determine that the rat infestation level corresponding to the preset target farmland is in a low-risk state. The preset first rat activity intensity index threshold is greater than or equal to the preset second rat activity intensity index threshold.

[0109] A first threshold for rodent activity intensity is preset, representing the minimum critical value for high rodent activity intensity. Higher rodent activity intensity corresponds to greater harm to the target farmland, and the target farmland is considered to be at high risk of rodent infestation. Similarly, a second threshold for rodent activity intensity is preset, representing the maximum critical value for low rodent activity intensity. Lower rodent activity intensity corresponds to less harm to the target farmland, and the target farmland is considered to be at low risk of rodent infestation. If the first and second thresholds for rodent activity intensity are not the same, a medium risk may exist between high and low risk.

[0110] Based on the above description, since the final determination is the target detection result corresponding to a single rat target detection, the rat activity intensity index corresponding to the preset target farmland is statistically analyzed based on several of the above final determinations. The rat activity intensity index includes, but is not limited to, rat population size and rat activity frequency. Rat population size refers to the number of rats detected. Rat activity frequency can be measured by the number of times rats appear within a certain period of time, such as the number of times rats appear in a day, and includes at least one of the following:

[0111] If the rat activity intensity index is greater than or equal to the preset first rat activity intensity index threshold, and if the above judgment is yes, the rat infestation level corresponding to the preset target farmland is determined to be in a high-risk state. Similarly, if the above judgment is no, the rat infestation level corresponding to the preset target farmland is determined not to be in a high-risk state.

[0112] The system determines whether the rat activity intensity index is less than a preset second rat activity intensity index threshold. If the above determination is yes, the rat infestation level corresponding to the preset target farmland is determined to be in a low-risk state. Similarly, if the above determination is no, the rat infestation level corresponding to the preset target farmland is determined to be not in a low-risk state. The preset first rat activity intensity index threshold is greater than or equal to the preset second rat activity intensity index threshold.

[0113] According to the embodiments of the present invention, by accurately judging whether the current monitoring image contains a rat image, the degree of harm caused by rats to farmland can be accurately assessed, and intelligent monitoring of farmland rats and their harm can be achieved automatically and efficiently, thereby improving the level of agricultural intelligence.

[0114] It should be noted that the YOLO-based methods for detecting the severity of rat diseases described in the above embodiments can be recombined as needed to obtain combined implementation schemes, but all are within the scope of protection claimed by this invention.

[0115] In one embodiment, a YOLO-based system for detecting the severity of rodent diseases is provided. This YOLO-based system corresponds one-to-one with the YOLO-based methods for detecting the severity of rodent diseases described in the previous embodiments. Please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic block diagram of a YOLO-based system for detecting the severity of rat diseases, provided in an embodiment of the present invention. Figure 5 As shown, the YOLO-based system 500 for detecting the severity of rat diseases includes a first determination module 501, a first judgment module 502, and a second judgment module 503. The detailed descriptions of each functional module are as follows:

[0116] The first determining module 501 is used to determine the current monitoring image and corresponding rat monitoring audio corresponding to the preset target farmland, and to determine the preset unique morphological characteristics of rats.

[0117] The first determination module 502 is used to determine the mouse-specific morphological features of the current frame corresponding to the current monitoring image when it is determined that the mouse monitoring audio contains mouse behavior sounds corresponding to mouse behavior, and based on the preset mouse-specific morphological features and the preset mouse target detection YOLO model, and when it is initially determined that the current monitoring image contains a mouse image.

[0118] The second determination module 503 is used to determine that the current monitoring image contains a rat image when the rat's unique morphological features in the current frame match the rat's movement morphological trajectory, and to determine the rat infestation level of the preset target farmland based on several of the above final determinations.

[0119] In one embodiment, the second determination module 503 includes:

[0120] The first determining submodule is used to determine several adjacent monitoring images based on time series corresponding to the current monitoring image;

[0121] The second determination submodule is used to determine the unique morphological features of mice in adjacent frames corresponding to the monitored adjacent images, and obtain several unique morphological features of mice in adjacent frames.

[0122] The sequence composition submodule is used to combine the unique morphological features of the mouse in the current frame with the unique morphological features of the mouse in all adjacent frames based on the corresponding time series, so as to obtain the unique morphological feature sequence of the mouse.

[0123] The first judgment submodule is used to determine whether the unique morphological feature sequence of the mouse conforms to the preset mouse action pattern.

[0124] The first determination submodule is used to determine, if the above determination is true, that the unique morphological features of the mouse in the current frame match the mouse's action morphological trajectory.

[0125] In one embodiment, the first determination module 502 includes:

[0126] The first detection submodule is used to detect whether the current monitoring image contains a mouse image based on the preset unique morphological features of mice and the current monitoring image, and based on the preset mouse target detection YOLO model.

[0127] The first identification submodule is used to identify whether the mouse monitoring audio contains the sound of mouse behavior corresponding to the mouse behavior based on the preset mouse behavior voiceprint recognition model.

[0128] The third determination submodule is used to determine the unique morphological features of the mouse in the current frame corresponding to the current monitoring image when the current monitoring image contains a mouse image and the mouse monitoring audio contains mouse behavioral sounds.

[0129] In one embodiment, the first detection submodule includes:

[0130] The fourth determination submodule is used to determine a preset set of unique morphological features of mice, wherein the preset set of unique morphological features of mice includes at least one of the following: a preset set of unique morphological features of mouse head, a preset set of unique morphological features of mouse trunk, and a preset set of unique morphological features of mouse tail.

[0131] The second detection submodule is used to detect whether the current monitored image contains preset unique morphological features of a mouse head, preset unique morphological features of a mouse trunk, or preset unique morphological features of a mouse tail, based on a preset mouse target detection YOLO model.

[0132] The second determination submodule is used to determine, if the above detection is true, that the current monitored image contains a mouse image;

[0133] The third determination submodule is used to determine that no mouse image was detected in the current monitoring image if the above detection is negative.

[0134] In one embodiment, the first identification submodule includes at least one of the following:

[0135] The third detection submodule is used to detect whether the mouse monitoring audio contains mouse squeaking sounds corresponding to mouse interaction behaviors;

[0136] The fourth detection submodule is used to detect whether the mouse monitoring audio contains the squealing sounds of a mouse that correspond to a behavior of being violated by the mouse;

[0137] The fifth detection submodule is used to detect whether the mouse monitoring audio contains the mouse gnawing sound corresponding to the mouse gnawing behavior;

[0138] The sixth detection submodule is used to detect whether the mouse monitoring audio contains the rubbing sounds corresponding to the mouse's actions;

[0139] The seventh detection submodule is used to detect whether the mouse monitoring audio contains sounds corresponding to the mouse's diseased behavior.

[0140] In one embodiment, the first determining module 501 includes:

[0141] The fifth determination submodule is used to determine several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland. The several types of current initial monitoring images include at least two of the following: visible light current initial monitoring image, near-infrared current initial monitoring image, and thermal infrared current initial monitoring image.

[0142] The stitching submodule is used to align and stitch all the current initial monitoring images to obtain the current monitoring image corresponding to the preset target farmland.

[0143] In one embodiment, the fifth determining submodule includes:

[0144] The first acquisition submodule is used to respond to different preset acquisition trigger events and, based on the corresponding preset video acquisition equipment, acquire several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland.

[0145] The sixth determination submodule is used to determine several single-frame monitoring video images corresponding to each type of mouse monitoring video;

[0146] The seventh determination submodule is used to determine the current initial monitoring image of each type corresponding to the preset target farmland from several single-frame monitoring video images of each type based on a preset image determination method and the same time sequence, so as to obtain several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland.

[0147] In one embodiment, the first acquisition submodule includes at least one of the following:

[0148] The second acquisition submodule is used to acquire several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland, based on preset visible light video acquisition equipment and preset near-infrared video acquisition equipment, when the preset acquisition trigger event corresponds to preset good light conditions.

[0149] The third acquisition submodule is used to acquire several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland, based on the preset RGB-NIR multispectral camera video acquisition device and the preset thermal infrared video acquisition device, when the preset acquisition trigger event corresponds to the preset low light conditions.

[0150] The fourth acquisition submodule is used to acquire several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland, based on preset thermal infrared video acquisition equipment and preset shortwave infrared video acquisition equipment, when the preset acquisition trigger event corresponds to preset no-light conditions.

[0151] In one embodiment, the second determination module 503 includes:

[0152] The statistics submodule is used to calculate the rat activity intensity index corresponding to the preset target farmland based on several of the above final judgments, and includes at least one of the following:

[0153] The second judgment submodule is used to determine whether the rat activity intensity index is greater than or equal to the preset first rat activity intensity index threshold, and if the above judgment is yes, the rat infestation level corresponding to the preset target farmland is determined to be in a high-risk state.

[0154] The third judgment submodule is used to determine whether the rat activity intensity index is less than the preset second rat activity intensity index threshold. If the above judgment is true, the rat infestation level corresponding to the preset target farmland is determined to be in a low-risk state. The preset first rat activity intensity index threshold is greater than or equal to the preset second rat activity intensity index threshold.

[0155] This invention provides a YOLO-based system for detecting the severity of rodent diseases. It determines the current monitoring image and corresponding rodent monitoring audio corresponding to a preset target farmland, and identifies preset unique morphological features of the rodents. If the rodent monitoring audio contains sounds corresponding to rodent behavior, and based on the preset unique morphological features and a preset YOLO model for rodent target detection, if the current monitoring image is initially determined to contain a rodent image, then the system determines the unique morphological features of the rodent in the current frame corresponding to the current monitoring image. If the unique morphological features of the rodent in the current frame match the rodent's movement trajectory, then the system ultimately determines that the current monitoring image contains a rodent image, and further determines the system based on several of the aforementioned final determinations. This method determines the degree of rodent infestation corresponding to a preset target farmland, thereby realizing a multi-dimensional, cross-modal parallel detection strategy that combines voiceprint recognition with target detection based on the unique morphological features of rats. Furthermore, it implements a progressive hierarchical detection strategy of "preliminary judgment - re-judgment verification." Compared to the single-dimensional target detection in traditional technologies, this method can perform multiple composite verifications of rat detection from different angles, effectively improving the accuracy of farmland rat detection. It adapts to the complex environmental background, indistinct features, and camouflage-like fur of rats in farmland, thus accurately assessing the harm rats cause to farmland, improving the accuracy of farmland rodent infestation detection, and achieving automatic and efficient intelligent monitoring of farmland rats and their damage.

[0156] For specific limitations regarding the YOLO-based system for detecting the severity of rodent diseases, please refer to the limitations of the YOLO-based method for detecting the severity of rodent diseases mentioned above, which will not be repeated here. Each module in the aforementioned YOLO-based system for detecting the severity of rodent diseases can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or they can be stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0157] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0158] The data collection in this embodiment of the invention complies with the requirements of relevant laws and regulations, such as China's Personal Information Protection Law, GDPR (General Data Protection Regulation of the European Union), or information security standards of other countries and regions.

[0159] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting the severity of rat diseases based on YOLO, characterized in that, include: Determine the current monitoring images and corresponding rat monitoring audio corresponding to the preset target farmland, and determine the preset unique morphological characteristics of the rats; If it is determined that the mouse monitoring audio contains mouse behavior sounds corresponding to mouse behavior, and based on the preset mouse-specific morphological features and the preset mouse target detection YOLO model, if it is initially determined that the current monitoring image contains a mouse image, the mouse-specific morphological features of the current frame corresponding to the current monitoring image are determined. If the unique morphological features of the rat in the current frame match the rat's movement morphological trajectory, it is finally determined that the current monitoring image contains a rat image, and based on several of the above final determinations, the degree of rat infestation corresponding to the preset target farmland is determined. Wherein, the step of determining that the unique morphological features of the mouse in the current frame match the mouse's movement morphological trajectory includes: Determine several adjacent monitoring images based on time series corresponding to the current monitoring image; Determine the unique morphological features of mice in adjacent frames corresponding to the monitored adjacent images to obtain several unique morphological features of mice in adjacent frames; Based on the corresponding time series, the mouse-specific morphological features of the current frame and the mouse-specific morphological features of all the adjacent frames are combined to form a sequence, thus obtaining the mouse-specific morphological feature sequence. Determine whether the mouse-specific morphological feature sequence conforms to the preset mouse movement pattern corresponding to the mouse; If the above judgment is true, it is determined that the unique morphological features of the mouse in the current frame match the mouse's movement morphological trajectory.

2. The method for detecting the severity of rat diseases based on YOLO as described in claim 1, characterized in that, The step of determining the mouse-specific morphological features of the current frame corresponding to the current monitoring image after initially determining that the current monitoring image contains a mouse image, based on the preset mouse-specific morphological features and a preset mouse target detection YOLO model, includes: Based on the preset unique morphological features of mice and the current monitoring image, and based on the preset mouse target detection YOLO model, it is determined whether the current monitoring image contains a mouse image; Based on a preset rat behavior voiceprint recognition model, it is determined whether the rat monitoring audio contains rat behavior sounds corresponding to rat behavior. If the current monitoring image is detected to contain a mouse image and the mouse monitoring audio is identified to contain the sound of the mouse's behavior, the unique morphological features of the mouse in the current frame corresponding to the current monitoring image are determined.

3. The method for detecting the severity of rat diseases based on YOLO as described in claim 2, characterized in that, The step of detecting whether the current monitoring image contains a mouse image based on the preset mouse-specific morphological features and the current monitoring image, and based on the preset mouse target detection YOLO model, includes: A preset set of unique morphological features of mice is determined, wherein the preset set of unique morphological features of mice includes at least one of the following: a preset set of unique morphological features of mouse head, a preset set of unique morphological features of mouse torso, and a preset set of unique morphological features of mouse tail. Based on the preset mouse target detection YOLO model, detect whether the current monitoring image contains the preset mouse head unique morphological features, the preset mouse torso unique morphological features, or the preset mouse tail unique morphological features; If the above detection is positive, it is determined that the currently monitored image contains an image of a mouse. If the above detection is negative, it is determined that no image of a mouse was detected in the currently monitored image.

4. The method for detecting the severity of rat diseases based on YOLO as described in claim 2, characterized in that, Based on a preset rat behavior voiceprint recognition model, identify whether the rat monitoring audio contains rat behavior sounds corresponding to rat behavior, including at least one of the following: Detect whether the mouse monitoring audio contains mouse squeaking sounds corresponding to mouse interactive behaviors; The detection function checks whether the audio from the mouse monitoring system contains the squealing sounds of a mouse corresponding to a perceived act of aggression. Detect whether the mouse monitoring audio contains the mouse gnawing sound corresponding to the mouse gnawing behavior; Detect whether the mouse monitoring audio contains the rubbing sounds corresponding to the mouse's actions; The detection function checks whether the audio recordings of the mouse monitoring contain sounds corresponding to the mouse's diseased behavior.

5. The method for detecting the severity of rat diseases based on YOLO as described in claim 1, characterized in that, The process of determining the current monitoring image corresponding to the preset target farmland includes: Determine several types of current initial monitoring images for different spectral bands corresponding to the preset target farmland. The several types of current initial monitoring images include at least two of the following: visible light current initial monitoring image, near-infrared current initial monitoring image, and thermal infrared current initial monitoring image. All the current initial monitoring images are aligned and stitched together to obtain the current monitoring image corresponding to the preset target farmland.

6. The method for detecting the severity of rat diseases based on YOLO as described in claim 5, characterized in that, The determination of several types of current initial monitoring images corresponding to different spectral bands of the preset target farmland includes: Responding to different preset acquisition trigger events, and based on the corresponding preset video acquisition devices, the system acquires several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland. Determine a number of single-frame monitoring video images corresponding to each type of mouse monitoring video; Based on a preset image determination method and the same time sequence, the current initial monitoring image corresponding to each type of the preset target farmland is determined from several single-frame monitoring video images of each type, thereby obtaining several types of current initial monitoring images of different spectral bands corresponding to the preset target farmland.

7. The method for detecting the severity of rat diseases based on YOLO as described in claim 6, characterized in that, The method involves responding to different preset acquisition trigger events and, based on corresponding preset video acquisition devices, acquiring several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland, including at least one of the following: When the preset acquisition trigger event corresponds to preset good lighting conditions, based on preset visible light video acquisition equipment and preset near-infrared video acquisition equipment, several types of rat monitoring videos in different spectral bands corresponding to the preset target farmland are acquired. When the preset acquisition trigger event corresponds to the preset low light conditions, based on the preset RGB-NIR multispectral camera video acquisition device and the preset thermal infrared video acquisition device, several types of rat monitoring videos corresponding to different spectral bands of the preset target farmland are acquired. When the preset acquisition trigger event corresponds to a preset no-light condition, based on a preset thermal infrared video acquisition device and a preset short-wave infrared video acquisition device, several types of rat monitoring videos of different spectral bands corresponding to the preset target farmland are acquired.

8. The method for detecting the severity of rat diseases based on YOLO as described in claim 1, characterized in that, The determination of the rodent infestation level corresponding to the preset target farmland based on several of the aforementioned final determinations includes: Based on the aforementioned final determinations, the rat activity intensity index corresponding to the preset target farmland is statistically analyzed, including at least one of the following: Determine whether the rat activity intensity index is greater than or equal to a preset first rat activity intensity index threshold, and if the above determination is yes, determine that the rat infestation level corresponding to the preset target farmland is in a high-risk state; Determine whether the rat activity intensity index is less than a preset second rat activity intensity index threshold. If the above determination is yes, determine that the rat infestation level corresponding to the preset target farmland is in a low-risk state, wherein the preset first rat activity intensity index threshold is greater than or equal to the preset second rat activity intensity index threshold.

9. A YOLO-based system for detecting the severity of rat diseases, characterized in that, include: The first determining module is used to determine the current monitoring image and corresponding rat monitoring audio corresponding to the preset target farmland, and to determine the preset unique morphological characteristics of the rats; The first determination module is used to determine the mouse-specific morphological features of the current frame corresponding to the current monitoring image when it is determined that the mouse monitoring audio contains mouse behavior sounds corresponding to mouse behavior, and based on the preset mouse-specific morphological features and the preset mouse target detection YOLO model, and when it is initially determined that the current monitoring image contains a mouse image. The second determination module is used to determine that the current monitoring image contains a rat image when the rat-specific morphological features of the current frame match the rat movement morphological trajectory, and to determine the rat infestation level of the preset target farmland based on several of the above final determinations. The second determination module includes: The first determining submodule is used to determine several adjacent monitoring images based on time series corresponding to the current monitoring image; The second determination submodule is used to determine the unique morphological features of mice in adjacent frames corresponding to the monitored adjacent images, and obtain several unique morphological features of mice in adjacent frames. The sequence composition submodule is used to combine the unique morphological features of the mouse in the current frame with the unique morphological features of the mouse in all adjacent frames based on the corresponding time series, so as to obtain the unique morphological feature sequence of the mouse. The first judgment submodule is used to determine whether the unique morphological feature sequence of the mouse conforms to the preset mouse action pattern. The first determination submodule is used to determine, if the above determination is true, that the unique morphological features of the mouse in the current frame match the mouse's action morphological trajectory.

Images