Target detection algorithm based on deep learning
By using a deep learning-based target detection algorithm, leveraging the YOLOv8 algorithm and feature fusion strategy, the problems of low efficiency, poor accuracy, and limited scope in monkey behavior detection were solved. This enabled real-time and accurate monitoring and data analysis of monkey behavior, allowing for the inference of behavioral intentions and expanding the observation range.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-03-13
AI Technical Summary
Current technologies for detecting monkey behavior rely on manual observation, which suffers from problems such as low efficiency, accuracy being greatly affected by subjective factors, inability to achieve continuous and real-time monitoring, difficulties in data storage and analysis, and limited observation range.
We employ a deep learning-based target detection algorithm, utilizing the YOLOv8 algorithm combined with specific loss function optimization and feature fusion strategies. Through steps such as image acquisition, data transmission, preprocessing, detection and recognition, classification and judgment, we achieve automated detection and analysis of monkey behavior.
It enables real-time and accurate detection of monkey behavior, expands the observation range, improves data collection efficiency and accuracy, provides continuous monitoring and data analysis capabilities, infers the intentions of monkey behavior, and provides strong support for research.
Smart Images
Figure CN121661325A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of target detection algorithms, and more particularly to a target detection algorithm based on deep learning. Background Technology
[0002] Current technologies for detecting monkey behavior, which rely primarily on manual observation and lack effective technical means, have the following significant drawbacks:
[0003] Low observation efficiency: Manually observing monkey behavior requires a significant amount of time and manpower. Researchers need to spend long periods in the wild or in specific observation environments, continuously monitoring the monkeys' every move. When observing multiple monkey groups or large areas, the limitations of human resources become apparent, making it difficult to comprehensively and promptly record the behavior of all monkeys. This can lead to the omission of much behavioral information and the inability to obtain complete behavioral data.
[0004] The accuracy of observations is greatly affected by subjective factors: different researchers may make different judgments and records of monkey behavior due to differences in their personal knowledge, experience, and perspectives. For example, different people may have different understandings and interpretations of some subtle behavioral manifestations of monkeys, which makes it difficult to guarantee the accuracy and consistency of the observation results, thus affecting subsequent analysis and research on monkey behavior.
[0005] The inability to achieve continuous and real-time monitoring: Human observation is limited by time and energy, making it difficult to monitor monkey behavior continuously for 24 hours. Monkey behavior cannot be recorded at night or when researchers are resting. Furthermore, when monkeys exhibit sudden and significant behaviors, human observation may fail to capture them in time, missing crucial research information. In contrast, technological means can achieve uninterrupted real-time monitoring, compensating for the time limitations of human observation.
[0006] Data storage and analysis are challenging: Most data recorded through manual observation is documented in paper or simple electronic files, requiring significant additional work for organization, storage, and analysis. As observation time and data volume increase, manual data processing becomes increasingly difficult, making it challenging to quickly and accurately extract valuable information from massive datasets. However, utilizing technological means can enable automated data collection, storage, and preliminary analysis, greatly improving the efficiency and accuracy of data processing.
[0007] Limited Observation Range: Human observation is typically limited to a confined area, such as a specific observation zone or point. It is difficult to comprehensively track large-scale monkey activities, such as migration and changes in foraging ranges. Furthermore, in areas with complex terrain and harsh environments, human observation is extremely difficult or even impossible, thus limiting a comprehensive understanding of monkey behavior. Summary of the Invention
[0008] The technical problem to be solved by this invention is to overcome the shortcomings of existing technologies, such as low observation efficiency, difficulty in data storage, and errors in analysis results, and to provide a target detection algorithm based on deep learning.
[0009] The present invention solves the above-mentioned technical problems through the following technical solution:
[0010] This invention provides a deep learning-based object detection algorithm, which includes the following steps:
[0011] Step 1: Image Acquisition. Acquire the original image of the region.
[0012] Step 2: Data transmission. The original image obtained in Step 1 is transmitted to the data processing unit.
[0013] Step 3: Image preprocessing. After receiving the image data, the data processing unit preprocesses the image.
[0014] Step 4: Detection and Recognition. The YOLOv8 algorithm is used to detect and recognize targets in the preprocessed image.
[0015] Step 5: Classification and Judgment. Based on the data detected in Step 4 and the pre-set data, classify and judge the data.
[0016] Step Six: Data Storage and Output. The data from the above steps is stored in the storage module, and the data is displayed and output through the display and output module.
[0017] In this technical solution, the image acquisition device acquires raw images and transmits them to the data processing unit through the data transmission module. The data processing unit performs a series of operations on the images, including preprocessing, monkey detection, posture recognition, and behavior judgment. Part of the processed data is stored in the storage module, and part is displayed to the user through the display and output module. All parts work closely together to realize the monkey behavior detection function based on YOLOv8, enabling the system to detect monkey behavior in real time and accurately, and providing valuable information for researchers and related staff.
[0018] Preferably, the image preprocessing in step three incorporates the characteristics of different environments in the detection scene.
[0019] In this technical solution, the target detection algorithm can automatically adjust the image preprocessing parameters based on factors such as the image's lighting conditions and background complexity.
[0020] Preferably, step five involves automatically changing and optimizing the judgment rules based on the collected data.
[0021] In this technical solution, as data accumulates, the mechanism can make the system's judgment of monkey behavior more accurate and in line with the actual situation, thereby improving the system's adaptability and long-term effectiveness.
[0022] Preferably, the YOLOv8 algorithm in step four incorporates specific loss function optimization and feature fusion strategies.
[0023] In this technical solution, by introducing specific loss function optimization and feature fusion strategies, the detection accuracy of monkey body joints is improved in complex environments (such as insufficient light, strong background interference, monkeys occluding each other, etc.), thereby more accurately judging the monkey's posture and providing a reliable foundation for subsequent behavior analysis.
[0024] Preferably, the YOLOv8 algorithm in step four constructs a deep learning-based monkey behavior intention analysis model.
[0025] In this technical solution, by learning from a large amount of monkey behavior data, the intentions behind the monkey's behavior, such as foraging intentions, social intentions, and risk avoidance intentions, can be inferred based on information such as the monkey's posture, action sequence, and environment.
[0026] Preferably, the detection model framework of the target detection algorithm includes an image acquisition device, a data transmission module, a data processing unit, a storage module, and a display and output module;
[0027] The image acquisition device is connected to the data transmission module, the data transmission module is connected to the data processing unit, both the image acquisition device and the data processing unit are connected to the storage module, and the data processing unit is connected to the display and output module.
[0028] The data processing unit is equipped with the YOLOv8 algorithm.
[0029] Preferably, the image acquisition device comprises one or more of a camera, an infrared sensor, and a lidar.
[0030] In this technical solution, the image acquisition devices are distributed in the monkey's activity area, which can collect image information of the monkey's activities in real time, providing a data foundation for subsequent analysis.
[0031] Preferably, the data transmission module is used to transmit the image data acquired by the image acquisition device to the data processing unit in real time, and the data transmission method adopts wireless transmission technology.
[0032] In this technical solution, the data transmission module is used to ensure that image data can reach the processing unit in a timely and accurate manner for subsequent analysis.
[0033] Preferably, the storage module is used to store the raw image data acquired by the image acquisition device and the processed data.
[0034] In this technical solution, the storage module is used to save data, provide researchers with historical data references, and can also be used for further optimization and training of the algorithm.
[0035] Preferably, the display and output module is used to notify and display the detection results in an intuitive way.
[0036] In this technical solution, the display and output module is used to enable users to easily obtain and understand the test results.
[0037] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of the present invention.
[0038] The positive and progressive effects of this invention are as follows:
[0039] In this invention, the image acquisition device acquires raw images and transmits them to the data processing unit through the data transmission module. The data processing unit performs a series of operations on the images, including preprocessing, monkey detection, posture recognition, and behavior judgment. Part of the processed data is stored in the storage module, and part is displayed to the user through the display and output module. All parts work closely together to realize the monkey behavior detection function based on YOLOv8, enabling the system to detect monkey behavior in real time and accurately, and providing valuable information for researchers and related staff.
[0040] It can quickly and comprehensively collect monkey behavioral data, improve the efficiency of data collection, avoid data omissions caused by the limitations of manual observation, and obtain more complete information on monkey behavior.
[0041] It can quickly and comprehensively collect monkey behavioral data, improve the efficiency of data collection, avoid data omissions caused by the limitations of manual observation, and obtain more complete information on monkey behavior.
[0042] Regardless of day or night, or inclement weather, the system can continuously monitor monkey behavior, promptly capture sudden behaviors, and provide continuous and complete data for studying monkey behavior patterns and rules.
[0043] The system enables automatic storage, efficient retrieval, and in-depth analysis of monkey behavior data, allowing for the rapid extraction of key information from large amounts of data, providing strong data support for researchers, and promoting the development of monkey behavior research.
[0044] Develop technologies that can adapt to different environments and expand the observation range, such as using drones and remote monitoring equipment, to track and monitor monkeys' activities over a wide area, breaking through the geographical limitations of manual observation and gaining a more comprehensive understanding of monkeys' ecological behavior.
[0045] To address the specific needs of monkey behavior detection, the YOLOv8 algorithm was customized and improved. By introducing specific loss function optimization and feature fusion strategies, the detection accuracy of monkey body joints was improved in complex environments (such as insufficient light, strong background interference, and monkeys occluding each other), thus more accurately judging the monkey's posture and providing a reliable foundation for subsequent behavior analysis. This improvement makes the model more accurate and robust than the traditional unmodified YOLOv8 algorithm when handling monkey behavior detection tasks.
[0046] Based on the YOLOv8 detection of monkey postures, a deep learning-based monkey behavior intention analysis model was further constructed. This model, through learning from a large amount of monkey behavior data, can infer the intentions behind monkey behavior, such as foraging intentions, social intentions, and risk avoidance intentions, based on information such as monkey posture, action sequence, and environment. This provides a new perspective and method for monkey behavior research.
[0047] By combining the characteristics of different environments in monkey behavior detection scenarios, the parameters of image preprocessing, such as brightness adjustment, contrast enhancement, and noise removal, are automatically adjusted according to factors such as image lighting conditions and background complexity, thereby improving the image quality input to the YOLOv8 algorithm and further enhancing the effect of monkey detection and pose recognition.
[0048] The system can automatically update and optimize the rules used to judge monkey behavior based on newly collected monkey behavior data. As data accumulates, this mechanism can make the system's judgment of monkey behavior more accurate and in line with the actual situation, thereby improving the system's adaptability and long-term effectiveness. Attached Figure Description
[0049] Figure 1 This is an overall flowchart of the deep learning-based target detection algorithm according to an embodiment of the present invention.
[0050] Figure 2 for Figure 1 The diagram shows the structural principle of a deep learning-based object detection algorithm.
[0051] Explanation of reference numerals in the attached figures
[0052] 1. Image acquisition device; 2. Data transmission module; 3. Data processing unit; 4. Storage module; 5. Display and output module. Detailed Implementation
[0053] The present invention will be further illustrated by way of embodiments below, but the present invention is not limited to the scope of the embodiments described herein.
[0054] Figures 1 to 2 The diagram shown is a structural schematic of an embodiment of the object detection algorithm based on deep learning of the present invention.
[0055] Structurally, YOLOv8 is a deep learning-based object detection algorithm that improves and optimizes upon the YOLO series of algorithms. The overall network structure of YOLOv8 comprises several key components.
[0056] First is the backbone network, which is responsible for feature extraction from the input image. The backbone network typically uses a convolutional neural network (CNN) structure, extracting low- to high-level features from the image through a series of convolutional layers, pooling layers, and other operations. These features form the basis for subsequent object detection and pose estimation.
[0057] Secondly, there's the neck network, which fuses and processes the features extracted by the backbone network, further enhancing the expressive power of the features. Through structures such as the Feature Pyramid Network (FPN), the neck network fuses features from different levels, enabling the model to detect targets and perform pose analysis at different scales.
[0058] Finally, there's the head network, a crucial component of YOLOv8 for object detection and pose detection. In pose detection, the head network outputs the positional information of various joints on the monkey's body, allowing the determination of the monkey's pose. For example, it detects the coordinates of joints in the monkey's limbs, head, and torso, thus accurately describing the monkey's movement and posture.
[0059] From a theoretical perspective, YOLOv8's pose detection is based on deep learning principles of object detection and pose estimation. During the training phase, the model learns using a large amount of image data with monkey pose annotations. The annotation data contains the accurate location information of various joints in the monkey's body. The model continuously adjusts its parameters through the backpropagation algorithm to minimize the error between the model's predictions and the annotation data.
[0060] During the inference phase, when an image containing a monkey is input, the backbone network first extracts features from the image, obtaining a series of feature maps. Then, the neck network fuses and processes these feature maps to enhance their expressive power. Finally, the head network predicts the positions of various joints on the monkey's body based on the processed feature maps. Using this joint position information, the specific behavior of the monkey can be determined, such as whether it is climbing, jumping, or eating.
[0061] YOLOv8's posture detection is highly efficient and accurate in monkey behavior detection, capable of quickly and accurately identifying various monkey postures, providing strong technical support for animal behavior research, animal protection, and other fields.
[0062] A deep learning-based object detection algorithm includes the following steps:
[0063] Step 1: Image Acquisition. Acquire the original image of the region.
[0064] Image acquisition device 1 acquires images of the monkey's activity area according to the settings, obtaining raw images containing the monkey's behavior. The acquired images should have a sufficiently high resolution to ensure that the monkey's posture and movement details can be clearly identified.
[0065] Step 2: Data transmission. The original image obtained in Step 1 is transmitted to the data processing unit 3.
[0066] The collected data is transmitted wirelessly to the data processing unit 3 in real time via the data transmission module 2. During the transmission process, the data is encrypted to ensure its security and integrity.
[0067] Step 3: Image preprocessing. After receiving the image data, the data processing unit 3 preprocesses the image.
[0068] Preprocessing includes resizing the image to meet the input requirements of the YOLOv8 algorithm; and performing image enhancement operations, such as brightness adjustment and contrast enhancement, to improve image quality and facilitate subsequent detection.
[0069] Step 4: Detection and Recognition. The YOLOv8 algorithm is used to detect and recognize targets in the preprocessed image.
[0070] The YOLOv8 algorithm is used to perform target detection on the preprocessed image to identify the monkey in the image. At the same time, the pose detection function of YOLOv8 is used to detect the position of various joints of the monkey's body, thereby determining the monkey's pose.
[0071] For example, by detecting the coordinates of joints in a monkey's limbs, head, etc., it can be determined whether the monkey is climbing, jumping, or resting.
[0072] Step 5: Classification and Judgment. Based on the data detected in Step 4 and the pre-set data, classify and judge the data.
[0073] Based on the detected monkey postures and pre-set behavioral rules, the monkey's behavior is classified and judged.
[0074] For example, if the monkey's limb joints are found to be in the position of climbing and its head is facing upwards, it is determined that the monkey is climbing.
[0075] These behavioral rules can be derived through the analysis and summarization of a large amount of monkey behavior data.
[0076] Step 6: Data storage and output. The data from the above steps is stored in storage module 4, and the data is displayed and output through display and output module 5.
[0077] The detected monkey behavior information and raw image data are stored in storage module 4. At the same time, the monkey behavior information is displayed and output through display and output module 5.
[0078] For example, the monkey's behavior can be displayed in real time on a monitoring screen, or information about abnormal behavior can be sent to relevant personnel.
[0079] In this technical solution, the image acquisition device acquires raw images and transmits them to the data processing unit through the data transmission module. The data processing unit performs a series of operations on the images, including preprocessing, monkey detection, posture recognition, and behavior judgment. Part of the processed data is stored in the storage module, and part is displayed to the user through the display and output module. All parts work closely together to realize the monkey behavior detection function based on YOLOv8, enabling the system to detect monkey behavior in real time and accurately, and providing valuable information for researchers and related staff.
[0080] This invention develops a highly efficient automated monkey behavior monitoring technology that can quickly and comprehensively collect monkey behavior data, improve data collection efficiency, avoid data omissions caused by the limitations of manual observation, and obtain more complete monkey behavior information.
[0081] Design objective and accurate monkey behavior detection algorithms and systems to reduce the impact of subjective differences in human observation on behavior judgment and ensure high consistency and accuracy in the detection and recording of monkey behavior under different environmental and observer conditions;
[0082] To build a monkey behavior monitoring system that enables continuous and real-time monitoring, the system can monitor monkey behavior without interruption under various conditions such as day and night and inclement weather, and promptly capture sudden behaviors of monkeys, providing continuous and complete data for studying monkey behavior patterns and rules.
[0083] Develop a technology platform that facilitates data storage, management and analysis, enabling automatic storage, efficient retrieval and in-depth analysis of monkey behavior data, and quickly extracting key information from large amounts of data, providing strong data support for researchers and promoting the development of monkey behavior research.
[0084] Develop technologies that can adapt to different environments and expand the observation range, such as using drones and remote monitoring equipment, to track and monitor monkeys' activities over a wide area, breaking through the geographical limitations of manual observation and gaining a more comprehensive understanding of monkeys' ecological behavior.
[0085] In step three, image preprocessing is combined with the characteristics of different environments in the detection scene.
[0086] In this technical solution, the target detection algorithm can automatically adjust the image preprocessing parameters based on factors such as the image's lighting conditions and background complexity.
[0087] Step five involves automatically changing and optimizing the judgment rules based on the collected data.
[0088] In this technical solution, as data accumulates, the mechanism can make the system's judgment of monkey behavior more accurate and in line with the actual situation, thereby improving the system's adaptability and long-term effectiveness.
[0089] The YOLOv8 algorithm in step four introduces specific loss function optimization and feature fusion strategies.
[0090] In this technical solution, by introducing specific loss function optimization and feature fusion strategies, the detection accuracy of monkey body joints is improved in complex environments (such as insufficient light, strong background interference, monkeys occluding each other, etc.), thereby more accurately judging the monkey's posture and providing a reliable foundation for subsequent behavior analysis.
[0091] The YOLOv8 algorithm in step four constructs a deep learning-based model for analyzing monkey behavior and intent.
[0092] In this technical solution, by learning from a large amount of monkey behavior data, the intentions behind the monkey's behavior, such as foraging intentions, social intentions, and risk avoidance intentions, can be inferred based on information such as the monkey's posture, action sequence, and environment.
[0093] The target detection algorithm's detection model framework includes an image acquisition device 1, a data transmission module 2, a data processing unit 3, a storage module 4, and a display and output module 5;
[0094] The image acquisition device 1 is connected to the data transmission module 2, the data transmission module 2 is connected to the data processing unit 3, both the image acquisition device 1 and the data processing unit 3 are connected to the storage module 4, and the data processing unit 3 is connected to the display and output module 5.
[0095] The data processing unit 3 is equipped with the YOLOv8 algorithm.
[0096] Data processing unit 3 is equipped with a monkey behavior detection algorithm based on YOLOv8;
[0097] Data processing unit 3 can be a high-performance server or embedded device, such as an NVIDIA Jetson series development board; it receives image data from the data transmission module, uses the YOLOv8 algorithm to detect monkeys in the images, and identifies their postures and behaviors;
[0098] The function of data processing unit 3 is to analyze and process the raw image data to obtain information about the monkey's behavior.
[0099] The image acquisition device 1 consists of one or more of a camera, an infrared sensor, and a lidar.
[0100] In this technical solution, the image acquisition device 1 is distributed in the monkey's activity area, which can collect image information of the monkey's activities in real time, providing a data basis for subsequent analysis.
[0101] The data transmission module 2 is used to transmit the image data acquired by the image acquisition device 1 to the data processing unit 3 in real time, and the data transmission method adopts wireless transmission technology.
[0102] In this technical solution, the data transmission module 2 is used to ensure that the image data can reach the processing unit in a timely and accurate manner for subsequent analysis.
[0103] The data transmission module 2 is responsible for transmitting the image data acquired by the image acquisition device 1 to the data processing unit in real time. It can use wireless transmission technology, such as Wi-Fi, 4G / 5G, etc., to ensure that the data can be transmitted quickly and stably. Its function is to ensure that the image data can reach the processing unit in a timely and accurate manner for subsequent analysis.
[0104] The storage module 4 is used to store the raw image data acquired by the image acquisition device 1 and the processed data.
[0105] In this technical solution, storage module 4 is used to save data, providing researchers with historical data references, and can also be used for further optimization and training of the algorithm.
[0106] Storage module 4 is used to store the collected raw image data and the processed monkey behavior data; it can use hard disk arrays or cloud storage to facilitate subsequent data query and analysis.
[0107] The function of storage module 4 is to save data, provide researchers with historical data references, and also be used for further optimization and training of the algorithm.
[0108] The display and output module 5 is used to notify and display the detection results in an intuitive way.
[0109] In this technical solution, the display and output module 5 is used to enable users to easily obtain and understand the test results.
[0110] The display and output module 5 presents the detection results in an intuitive way, such as displaying the monkey's real-time behavior information on the monitoring screen, or presenting statistical data of the monkey's behavior in the form of charts;
[0111] The test results can also be output to other devices or systems, such as sending text messages to relevant personnel to notify them of abnormal monkey behavior;
[0112] The purpose of the display and output module 5 is to enable users to easily obtain and understand the test results.
[0113] To address the specific needs of monkey behavior detection, the YOLOv8 algorithm was customized and improved. By introducing specific loss function optimization and feature fusion strategies, the detection accuracy of monkey body joints was improved in complex environments (such as insufficient light, strong background interference, and monkeys occluding each other), thus more accurately judging the monkey's posture and providing a reliable foundation for subsequent behavior analysis. This improvement makes the model more accurate and robust than the traditional unmodified YOLOv8 algorithm when handling monkey behavior detection tasks.
[0114] Based on the YOLOv8 detection of monkey postures, a deep learning-based monkey behavior intention analysis model was further constructed. This model, through learning from a large amount of monkey behavior data, can infer the intentions behind monkey behavior, such as foraging intentions, social intentions, and risk avoidance intentions, based on information such as monkey posture, action sequence, and environment. This provides a new perspective and method for monkey behavior research.
[0115] By combining the characteristics of different environments in monkey behavior detection scenarios, the parameters of image preprocessing, such as brightness adjustment, contrast enhancement, and noise removal, are automatically adjusted according to factors such as image lighting conditions and background complexity, thereby improving the image quality input to the YOLOv8 algorithm and further enhancing the effect of monkey detection and pose recognition.
[0116] The system can automatically update and optimize the rules used to judge monkey behavior based on newly collected monkey behavior data. As data accumulates, this mechanism can make the system's judgment of monkey behavior more accurate and in line with the actual situation, thereby improving the system's adaptability and long-term effectiveness.
[0117] Example 1 of image acquisition device 1
[0118] Image acquisition device 1 uses high-definition cameras, which are distributed in the monkey's activity area, such as forests and monkey enclosures in zoos. These cameras can collect image information of monkey activities in real time, providing a data basis for subsequent analysis.
[0119] The camera needs to have a certain night vision capability to ensure that it can acquire clear images under different lighting conditions;
[0120] The camera's purpose is to acquire raw image data containing monkey behavior.
[0121] Example 2 of image acquisition device 1
[0122] As one embodiment of this application, its difference from other embodiments is that the image acquisition device 1 uses an infrared sensor and a camera fusion;
[0123] Infrared sensors are deployed in the monkey activity area, combined with cameras; the infrared sensors can detect the monkeys' thermal signals in real time, obtaining their approximate location and movement information; the cameras are used to capture image details of the monkeys.
[0124] The data fusion algorithm is used to fuse data from infrared sensors and image data from cameras.
[0125] When performing behavior detection, the approximate location of the monkey is first determined based on the data from the infrared sensor. Then, the camera focuses on the area to obtain a clear image. Finally, image processing algorithms (which can be the deep learning or traditional methods mentioned above) are used to detect and judge the monkey's posture and behavior.
[0126] Infrared sensors have advantages in nighttime or low-light environments. When combined with cameras, they can compensate for the shortcomings of cameras in specific environments and improve the stability and accuracy of monkey behavior detection.
[0127] Example 3 of image acquisition device 1
[0128] As one embodiment of this application, its difference from other embodiments is that the image acquisition device 1 integrates the lidar and the camera;
[0129] The LiDAR was used to acquire 3D point cloud data of the monkey's activity area, while the camera acquired 2D image data.
[0130] By using a point cloud and image registration algorithm, the two data are fused. Using the fused data, the monkey's position and posture can be determined more accurately. In addition, the monkey's behavior can be detected by combining deep learning or traditional object detection and posture estimation methods.
[0131] For example, by analyzing the spatial relationships of different parts of a monkey's body in point cloud data and combining this with appearance information in images, we can determine the monkey's behavior.
[0132] The three-dimensional information provided by LiDAR can offer richer dimensions for monkey behavior detection. When fused with camera data, it can more accurately detect monkey behavior in complex environments.
[0133] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.
Claims
1. A deep learning-based object detection algorithm, characterized in that: The target detection algorithm includes the following steps: Step 1: Image Acquisition. Acquire the original image of the region. Step 2: Data transmission, the original image obtained in step 1 is transmitted to the data processing unit (3); Step 3: Image preprocessing. After receiving the image data, the data processing unit (3) preprocesses the image. Step 4: Detection and Recognition. The YOLOv8 algorithm is used to detect and recognize targets in the preprocessed image. Step 5: Classification and Judgment. Based on the data detected in Step 4 and the pre-set data, classify and judge the data. Step 6: Data storage and output. The data from the above steps is stored in the storage module (4), and the data is displayed and output through the display and output module (5).
2. The deep learning-based target detection algorithm as described in claim 1, characterized in that: In step three, image preprocessing is combined with the characteristics of different environments in the detection scene.
3. The deep learning-based target detection algorithm as described in claim 1, characterized in that: Step five involves automatically changing and optimizing the judgment rules based on the collected data.
4. The deep learning-based target detection algorithm as described in claim 1, characterized in that: The YOLOv8 algorithm in step four introduces specific loss function optimization and feature fusion strategies.
5. The deep learning-based target detection algorithm as described in claim 4, characterized in that: The YOLOv8 algorithm in step four constructs a deep learning-based model for analyzing monkey behavior and intent.
6. The deep learning-based target detection algorithm as described in claim 1, characterized in that: The target detection algorithm's detection model framework includes an image acquisition device (1), a data transmission module (2), a data processing unit (3), a storage module (4), and a display and output module (5); The image acquisition device (1) is connected to the data transmission module (2), the data transmission module (2) is connected to the data processing unit (3), the image acquisition device (1) and the data processing unit (3) are both connected to the storage module (4), and the data processing unit (3) is connected to the display and output module (5). The data processing unit (3) is equipped with the YOLOv8 algorithm.
7. The deep learning-based target detection algorithm as described in claim 6, characterized in that: The image acquisition device (1) consists of one or more of a camera, an infrared sensor, and a lidar.
8. The deep learning-based target detection algorithm as described in claim 6, characterized in that: The data transmission module (2) is used to transmit the image data acquired by the image acquisition device (1) to the data processing unit (3) in real time. The data transmission method adopts wireless transmission technology.
9. The deep learning-based target detection algorithm as described in claim 6, characterized in that: The storage module (4) is used to store the original image data and the processed data acquired by the image acquisition device (1).
10. The deep learning-based target detection algorithm as described in claim 6, characterized in that: The display and output module (5) is used to notify and display the detection results in an intuitive way.