Inspection tool suitable for preventing small animals from invading in complex environment

Abstract

The invention discloses an inspection tool for preventing small animals from invading in a complex environment, relates to the technical field of electrical equipment maintenance, and aims at comprehensively inspecting and monitoring fireproof and plugging parts so as to restrain potential safety hazards. The tool comprises an insulating operating rod and an industrial steering endoscope, one end of the operating rod is connected with a camera, and safety is guaranteed through insulating design. The inspection tool can be flexibly operated in a narrow space so as to observe areas which are difficult to touch, and real-time monitoring of small animal invasion behaviors is ensured. In the inspection process, by means of the display screen of the endoscope host, an operator can observe and record the invasion signs of the small animals in real time, and important pictures are stored through the video recording function. The endoscope host is provided with an artificial intelligence image recognition function, real-time analysis is carried out on collected images based on a convolutional neural network, and small animal type recognition and intrusion judgment are achieved. According to the invention, the efficiency and accuracy of small animal intrusion detection are improved.

Description

Technical Field

[0001] The invention relates to the technical field of electrical equipment maintenance, and in particular to an inspection tool suitable for preventing small animals from intruding in complex environments. Background Art

[0002] In modern power facilities such as substations and switch stations, efficient and safe transmission of electric energy is the key to ensuring the stable operation of the power system. To achieve this goal, the power plant is usually equipped with important functional areas such as high-voltage rooms, main control rooms, and protection rooms. The cable configuration in these areas is particularly important, involving the reasonable layout and sealing of primary and secondary cables. At present, most power facilities use primary and secondary cables for power transmission and communication.

[0003] However, the entry and exit of cables are important points to prevent small animals from invading the equipment area. Once these blocking measures fail, small animals may enter the equipment area, causing serious consequences, such as equipment short circuit failure or gnawing on important equipment such as communication cables, thus causing equipment shutdown, system failure or safety accidents.

[0004] In new construction, expansion, production overhaul and technical transformation, or new interval construction by customers, the insertion of cables will inevitably damage the existing plugging. In addition, the inlets and outlets of cables are often densely distributed, and the light is dim and the cable trench space is limited, which brings great difficulties to daily plugging inspection and project acceptance. This makes it difficult for operators to find plugging holes under the cable layer and in dead corners in time during inspection, thus burying safety hazards. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an anti-small animal intrusion inspection tool suitable for use in complex environments. The anti-small animal intrusion inspection tool can inspect, accept and retain current data on fire prevention and anti-small animal blocking areas in all directions without blind spots, thereby curbing safety hazards at the source stage.

[0006] The technical solution adopted by the present invention is:

[0007] A tool for preventing small animals from intruding in complex environments, the tool comprising an insulating operating rod and an industrial steering endoscope; a wire clamp is installed at one end of the insulating operating rod, and the other end is a handle; the industrial steering endoscope comprises an endoscope host, and a camera connected to the endoscope host by wire; the wire clamp clamps the connection line between the endoscope host and the camera, and the clamping position is close to one end of the camera; the tool for preventing small animals from intruding in complex environments moves the camera of the industrial steering endoscope into a narrow space through the insulating operating rod, and inspects small animal intrusion through the host of the industrial steering endoscope.

[0008] Furthermore, the insulating operating rod includes a plurality of rod sections, and each rod section is connected through a steering positioning joint.

[0009] Furthermore, the endoscope host includes a display screen and a display screen mounting seat; an electric steering mechanism is installed at one end of the camera close to the connecting cable, and a lighting component is also provided inside the camera in addition to the camera component; the control switch of the electric steering mechanism is installed on the display screen mounting seat; the connecting cable is a Type-C data cable.

[0010] A method for inspecting small animals for intrusion in a complex environment, based on the above-mentioned small animal intrusion prevention inspection tool applicable to a complex environment, comprises the following steps:

[0011] Step 1: Connect the equipment: Connect the industrial steering endoscope to the insulating operating rod, and connect it to the endoscope host with a Type-C data cable; turn on the endoscope host and check whether the equipment connection is normal;

[0012] Step 2 Enter the inspection area: Carefully enter the inspection area according to the confirmed entry point; pay attention to ensure the safety of the surrounding environment and avoid potential electrical hazards;

[0013] Step 3: Introduce the endoscope: Hold the insulating operating rod and slowly insert the connected endoscope into the inspection area, trying to avoid collision between the camera and surrounding objects; use the length and flexibility of the insulating operating rod to insert the camera into corners, gaps or pipes that are difficult to observe;

[0014] Step 4: Image observation and adjustment: Observe the image transmitted by the camera in real time on the display screen of the endoscope host; use the operation buttons to adjust the brightness, contrast and magnification of the image to ensure clear observation; slowly move the insulating operating rod to adjust the angle and position of the endoscope, and comprehensively scan the inspection area, especially the channels and gaps where small animals may enter;

[0015] Step 5 Record and analyze: Look for signs of small animal intrusion, such as hair, feces, and bite marks; if suspicious signs are found, further zoom in on the image and observe in detail; use the host's video recording function to record important images for subsequent analysis and archiving;

[0016] Furthermore, the endoscope host can perform artificial intelligence image recognition, which processes and analyzes the images collected by the endoscope in real time through a deep learning algorithm to realize small animal type recognition and intrusion situation determination; the specific steps are as follows:

[0017] Step 501, data acquisition and preprocessing: the camera acquires real-time video stream, and the endoscope host receives image data; the image is preprocessed, and the preprocessing includes: denoising, normalization, cropping and enhancement to ensure that the image quality is suitable for model processing;

[0018] Step 502, model training: constructing a classification model based on a convolutional neural network (CNN), and using a data set containing a variety of small animal categories and background noise for training;

[0019] Step 503, real-time image analysis: the collected images are input into the trained model, features are extracted and classified; the recognition results are displayed in real time on the display screen of the endoscope host, and synchronously transmitted to the background host for storage and further analysis;

[0020] Step 504, alarm triggering and response: judging whether there is any small animal intrusion behavior based on the model recognition result; if intrusion is detected, triggering sound alarm, visual alarm and background notification;

[0021] Step 505, data update and optimization: by collecting data subsequently, the model is continuously updated to improve the recognition accuracy.

[0022] Furthermore, in step 501, when the endoscope host pre-processes the image, the image is first gray-scaled to reduce the computational complexity;

[0023] The calculation formula for grayscale is:

[0024] I gray =0.299R+0.587G+0.114B

[0025] In the formula, I gray Represents the pixel value of the grayscale image; R represents the pixel value of the red channel; G represents the pixel value of the green channel; B represents the pixel value of the blue channel;

[0026] The endoscope host normalizes the image and maps the pixel values ​​to the interval [0, 1];

[0027] The normalized calculation formula is:

[0028]

[0029] In the formula, I norm represents the normalized pixel value; I represents the pixel value to be normalized;

[0030] Finally, the endoscope host crops and scales the image to unify the image size.

[0031] Further, in step 502, the model structure of the convolutional neural network CNN classification model includes: a convolution layer for extracting low-level features, a multi-layer convolution and pooling layer for extracting high-level semantic features, and a fully connected layer for outputting classification results;

[0032] Among them, the calculation formula of the feature extraction process is as follows:

[0033] F=ReLu(W*I norm +b)

[0034] Where F is the feature map; W is the weight matrix of the convolution kernel; b is the bias term corresponding to each convolution kernel; * is the convolution operation; I norm Represents the normalized pixel value; ReLu(x)=max(0,x): is the activation function;

[0035] The calculation formula of the classification process is as follows:

[0036]

[0037] Where P(c) is the probability of category c; Z c is the output score of category c; exp is the mathematical exponential operation; It represents the sum of the output scores of all categories after being processed by the exponential function, n is the number of all possible categories, and i represents each category.

[0038] Further, in step 504, if the probability P(c) of a certain type of small animal in the classification result is greater than T; T is the set threshold, T=0.8, an alarm is triggered; the alarm mode includes: the screen highlights the intrusion target location, sound prompts, and data synchronization to the background host.

[0039] Further, in step 502, the model training process is as follows:

[0040] Data preparation: Data set D = {(X i ,y i )} includes images of various small animals and their labels; D represents a set containing multiple samples; X i ,y i Represents the features and labels of the i-th sample in the dataset; performs data augmentation by rotation, flipping, and scaling to increase sample diversity;

[0041] Model training: The loss function uses intersection and entropy loss:

[0042]

[0043] In the formula, L represents the average cross entropy loss of the entire data set; m is the number of samples; n is the number of categories; y ijis the true category label of sample i; y ij is the probability predicted by the model; log is usually the natural logarithm;

[0044] Optimize using the Adam optimizer:

[0045]

[0046] Where W represents the weight matrix in the neural network or model; b is the bias term corresponding to each convolution kernel; η is the learning rate; Represents the gradient of the loss function L with respect to the model parameters, including weights W and bias b;

[0047] In step 503, the collected images are input into the trained model to extract features and classify them. Specifically, the endoscope host performs reasoning on each frame of the image by preprocessing the image; inputs the model and outputs the category probability; and finally determines whether to trigger an alarm based on the result.

[0048] The beneficial effects of the present invention are:

[0049] 1. Comprehensive inspection capability: The tool is capable of conducting all-round, no-dead-angle inspections in complex environments, especially in small, hard-to-reach areas, ensuring comprehensive coverage of potential small animal intrusion points.

[0050] 2. High safety: The tool adopts an insulated operating rod, which increases the safety during operation, especially in electrical environments or other potential hazards, the operator can effectively avoid the risk of electric shock.

[0051] 3. Real-time image transmission and analysis: The industrial steering endoscope can observe and transmit images in real time, which can quickly identify the signs of intrusion of small animals and facilitate timely measures. At the same time, the endoscope host has artificial intelligence image recognition function, which can efficiently analyze and identify small animals and improve the accuracy and efficiency of inspection.

[0052] 4. Record and preserve evidence: The device supports video recording, which can record important scenes during the inspection process, facilitate subsequent analysis and archiving, and provide a basis for subsequent preventive measures.

[0053] 5. Reduce labor costs: Through the intelligent characteristics of this tool, you can reduce reliance on manual inspection, save labor costs, and improve the efficiency and accuracy of inspections.

[0054] 6. Early warning system: The tool’s built-in alarm function can promptly alarm when small animals invade, providing rapid response capabilities and reducing potential losses or risks.

[0055] 7. Data update and model optimization: With continued use, the tool can update the model through continuously collected data, improve recognition accuracy, and continuously improve the long-term effectiveness of the tool.

[0056] 8. Easy to operate: Operators can easily use the tool even in complex environments, which improves work efficiency and reduces operating difficulty.

[0057] In summary, this anti-small animal intrusion inspection tool suitable for complex environments not only improves the detection efficiency of small animal intrusion, but also enhances the ability to identify potential risks through intelligent means, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0059] Figure 1 This is a schematic diagram of the overall structure of the small animal intrusion prevention inspection tool of the present invention which is applicable to complex environments;

[0060] Figure 2 and Figure 3 It is a schematic diagram of the overall structure of the industrial steering endoscope of the present invention;

[0061] Figure 4 It is a structural schematic diagram of a display screen of an industrial steering endoscope of the present invention;

[0062] Figure 5 is a flow chart of a method for inspecting small animals for intrusion in a complex environment according to the present invention;

[0063] Figure 6 is a flow chart of artificial intelligence image recognition of the present invention;

[0064] In the figure: 1. Insulated operating lever; 2. Industrial steering endoscope; 3. Wire clamp; 4. Handle end; 5. Steering positioning joint; 6. Endoscope host; 7. Camera; 8. Connecting line; 9. Display screen; 10. Display screen mounting base; 11. Control switch. DETAILED DESCRIPTION

[0065] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0066] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents some embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should be noted that the embodiments of the present invention and the features and technical solutions in the embodiments can be combined with each other without conflict. It should be noted that similar numbers and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0067] Example 1

[0068] In order to solve the problem of small animal intrusion inspection in small areas of primary and secondary cables such as high-voltage rooms, main control rooms, and protection rooms, this embodiment provides an anti-small animal intrusion inspection tool suitable for complex environments. Figure 1 As shown, the small animal intrusion prevention inspection tool suitable for complex environments includes an insulating operating rod 1 and an industrial steering endoscope 2; wherein a wire clamp 3 is installed at one end of the insulating operating rod 1, and a handle end 4 is installed at the other end; Figure 1 , Figure 2 and Figure 3 As shown, the industrial steering endoscope 2 includes an endoscope host 6 and a camera 7 connected to the endoscope host 6 by wire; the wire clamp 3 at one end of the insulating operating rod 1 clamps the connecting line 8 between the endoscope host 6 and the camera 7, and the clamping point is close to one end of the camera 7.

[0069] The insulating operating rod 1 is used for the industrial steering endoscope 2, and protects the user from electric shock. The user can hold and flexibly operate the camera 7 of the industrial steering endoscope 2 to extend to a specified position through the insulating operating rod 1, so as to perform inspection at a certain safe distance in a complex environment. The wire clamp 3 is used as a connector to stably install the connection line 8 and the camera 7 of the industrial steering endoscope 2 on the insulating operating rod 1.

[0070] Further, as a preferred technical solution of this embodiment, Figure 1 As shown, the insulating operating rod 1 in this embodiment includes a plurality of rod sections, each of which is connected by a steering positioning joint 5. The insulating operating rod 1 can be deformed according to a narrow space through the steering positioning joint 5, thereby meeting the requirements of small animal intrusion inspection in a complex environment.

[0071] like Figure 2 , Figure 3 and Figure 4As shown, the industrial steering endoscope 2 is used for image acquisition and visual inspection of narrow spaces. It obtains real-time images of the inspected area through the camera 7 and displays them through the endoscope host 6. The endoscope host 6 includes a display screen 9 and a display screen mounting seat 10. An electric steering mechanism is installed at one end of the camera 7 close to the connecting cable. In addition to the camera component, the camera 7 is also provided with a lighting component; the control switch 11 of the electric steering mechanism is installed on the display screen mounting seat 10. The connecting cable is a Type-C data cable.

[0072] The camera 7 is equipped with a lighting component to ensure clear observation in low-light environments; the user can view the image in a small space in real time through the display screen 9. The design of the host makes it easy to carry and operate. The camera 7 is equipped with an electric steering mechanism near one end of the connecting cable, which can adjust the viewing angle in a small space to obtain the best field of view. In addition to the camera component, the camera 7 is also equipped with a lighting component to ensure clear imaging in dim environments. The electric steering mechanism improves the flexibility and controllability of the equipment, and the user can more easily control the direction of the camera 7 to adapt to complex and narrow inspection environments. The design of the display screen 9 makes it easy for the user to observe the image, and the mounting base ensures the stability and position adjustment of the display screen 9 during operation. The Type-C data cable ensures that the image data collected by the camera 7 can be quickly and effectively transmitted to the endoscope host 6 for easy processing and display.

[0073] In summary, the anti-small animal intrusion inspection tool suitable for complex environments moves the camera 7 of the industrial steering endoscope 2 into a narrow space through the insulating operating rod 1, and inspects the intrusion of small animals through the host of the industrial steering endoscope 2.

[0074] Example 2

[0075] Based on the small animal intrusion prevention inspection tool suitable for complex environments provided in Example 1, this embodiment provides a small animal intrusion inspection method in a complex environment, such as Figure 5 As shown, the small animal intrusion inspection method in the complex environment includes the following steps:

[0076] Step 1: Connect the equipment: Connect the industrial steering endoscope to the insulating operating rod, and connect it to the endoscope host with a Type-C data cable; turn on the endoscope host and check whether the equipment connection is normal.

[0077] Step 2 Enter the inspection area: Carefully enter the inspection area based on the confirmed entry point; pay attention to ensure the safety of the surrounding environment and avoid potential electrical hazards.

[0078] Step 3: Introduce the endoscope: Hold the insulating operating rod and slowly insert the connected endoscope into the inspection area, trying to avoid collision between the camera and surrounding objects; use the length and flexibility of the insulating operating rod to insert the camera into corners, gaps or pipes that are difficult to observe.

[0079] Step 4: Image observation and adjustment: Observe the image transmitted by the camera in real time on the display screen of the endoscope host; use the operating buttons to adjust the brightness, contrast and magnification of the image to ensure clear observation; slowly move the insulating operating rod to adjust the angle and position of the endoscope, and comprehensively scan the inspection area, especially the channels and gaps where small animals may enter.

[0080] Step 5 Record and Analyze: Look for signs of small animal intrusion, such as hair, feces, and bite marks. If any suspicious signs are found, further zoom in on the image and observe in detail. Use the host's video recording function to record important images for subsequent analysis and archiving.

[0081] This small animal intrusion inspection method in complex environments uses an insulated operating rod and an industrial steering endoscope, allowing inspectors to work safely in potentially electrical hazardous environments and reduce the risk of electric shock. The flexibility and high-resolution images of the endoscope can penetrate into hard-to-reach areas, making signs of small animal intrusion (such as hair, feces, etc.) more clearly visible, improving the comprehensiveness and accuracy of the inspection. Through the display screen of the endoscope host, inspectors can monitor the progress of the inspection in real time, adjust image parameters at any time, and optimize the observation effect, so as to discover potential problems in time. The introduction of the video recording function enables important images to be recorded, which is convenient for subsequent analysis and archiving, and helps to formulate targeted protective measures and improve long-term management capabilities. This method can adapt to a variety of complex environments, such as small spaces and hidden corners, to ensure no-dead-angle inspections and effectively prevent small animals from affecting facilities and equipment. Overall, this small animal intrusion inspection method in complex environments effectively improves the safety and effectiveness of small animal intrusion inspections, and provides reliable protection for the management of small animal intrusion prevention in substations, switch stations, etc.

[0082] Example 3

[0083] Based on Example 2, it is optimized; specifically, the endoscope host described in this embodiment can perform artificial intelligence image recognition, which uses a deep learning algorithm to process and analyze the images collected by the endoscope in real time to achieve small animal type recognition and intrusion situation determination; Figure 6 As shown, the specific steps are as follows:

[0084] Step 501, data acquisition and preprocessing: the camera acquires real-time video stream, and the endoscope host receives image data; the image is preprocessed, and the preprocessing includes: denoising, normalization, cropping and enhancement to ensure that the image quality is suitable for model processing.

[0085] When the endoscope host preprocesses the image, it first grayscales the image to reduce the computational complexity;

[0086] The calculation formula for grayscale is:

[0087] I gray =0.299R+0.587G+0.114B

[0088] In the formula, I gray Represents the pixel value of the grayscale image; R represents the pixel value of the red channel; G represents the pixel value of the green channel; B represents the pixel value of the blue channel;

[0089] The endoscope host normalizes the image and maps the pixel values ​​to the interval [0, 1];

[0090] The normalized calculation formula is:

[0091]

[0092] In the formula, I norm represents the normalized pixel value; I represents the pixel value to be normalized;

[0093] Finally, the endoscope host crops and scales the image to unify the image size.

[0094] Step 502, model training: construct a classification model based on a convolutional neural network (CNN), and use a data set containing multiple small animal categories and background noise for training.

[0095] The model structure of the convolutional neural network CNN classification model includes: a convolution layer for extracting low-level features, a multi-layer convolution and pooling layer for extracting high-level semantic features, and a fully connected layer for outputting classification results;

[0096] Among them, the calculation formula of the feature extraction process is as follows:

[0097] F=ReLu(W*I norm +b)

[0098] Where F is the feature map; W is the weight matrix of the convolution kernel; b is the bias term corresponding to each convolution kernel; * is the convolution operation; I norm Represents the normalized pixel value; ReLu(x)=max(0,x): is the activation function;

[0099] The calculation formula of the classification process is as follows:

[0100]

[0101] Where P(c) is the probability of category c; Z cis the output score of category c; exp is the mathematical exponential operation; It represents the sum of the output scores of all categories after being processed by the exponential function, n is the number of all possible categories, and i represents each category.

[0102] The model training process is as follows:

[0103] Data preparation: Data set D = {(X i ,y i )} includes images of various small animals and their labels; D represents a set containing multiple samples; X i ,y i Represents the features and labels of the i-th sample in the dataset; performs data augmentation by rotation, flipping, and scaling to increase sample diversity;

[0104] Model training: The loss function uses intersection and entropy loss:

[0105]

[0106] Where L represents the average cross entropy loss of the entire data set; m is the number of samples; n is the number of categories; y ij is the true category label of sample i; y ij is the probability predicted by the model; log is usually the natural logarithm;

[0107] Optimize using the Adam optimizer:

[0108]

[0109] Where W represents the weight matrix in the neural network or model; b is the bias term corresponding to each convolution kernel; η is the learning rate; Represents the gradient of the loss function L with respect to the model parameters, including weights W and bias b.

[0110] Step 503, real-time image analysis: the collected images are input into the trained model, features are extracted and classified; the recognition results are displayed in real time on the display screen of the endoscope host and synchronously transmitted to the background host for storage and further analysis.

[0111] The collected images are input into the trained model to extract features and classify them. Specifically, the endoscope host performs inference on each frame of the image by preprocessing the image; inputs the model and outputs the category probability; and finally determines whether to trigger an alarm based on the result.

[0112] Step 504, alarm triggering and response: based on the results of model recognition, determine whether there is any small animal intrusion behavior; if intrusion is detected, trigger sound alarm, visual alarm and background notification.

[0113] If the probability of a certain type of small animal in the classification result P(c)>T; T is the set threshold, T=0.8, then an alarm is triggered; the alarm methods include: the screen highlights the intrusion target location, sound prompts, and data synchronization to the background host.

[0114] Step 505, data update and optimization: by collecting data subsequently, the model is continuously updated to improve the recognition accuracy.

[0115] The artificial intelligence image recognition system carried by the above-mentioned endoscope host can significantly improve the effect of small animal type identification and intrusion determination. In terms of real-time and efficiency: the system can process the video stream collected by the endoscope in real time, quickly perform image recognition and output results, so that doctors or operators can grasp the situation in time; through automated image analysis, the need for manual intervention is reduced and work efficiency is improved. In terms of precision and accuracy: the classification model based on the convolutional neural network CNN can extract high-level features and provide higher recognition accuracy. This makes the identification of different small animals more accurate and reduces the false alarm rate. Through subsequent data updates, the model can continuously optimize and improve the recognition ability to ensure that the accuracy of the system continues to increase over time. In terms of decision support: once the intrusion behavior of small animals is identified, the system can immediately trigger sound and light alarms and background notifications to alert operators in time and facilitate rapid response. The synchronously transmitted recognition results can be stored in the background for subsequent analysis and evaluation, providing data support for decision-making. In terms of image processing effect: through denoising, normalization, cropping and enhancement steps, the image quality of the input model is ensured and the robustness of model processing is improved. Standardizing images (such as grayscale and unifying the size) makes it easier for the model to learn and adapt, and improves the consistency of processing.

[0116] In summary, this anti-small animal intrusion inspection tool suitable for complex environments is equipped with a human-machine intelligent image recognition system, which not only improves the real-time and accuracy of recognition, but also brings significant effect improvement to practical applications. This has positive significance for improving work efficiency, reducing human errors, and enhancing monitoring and safety capabilities.

[0117] At the same time, it should be pointed out that the above description is only a typical embodiment of the present invention and does not limit the present invention in any form. Any changes made to the above embodiments based on the ideas and technical essence of the present invention are within the protection scope of the present invention.

Claims

1. A small animal intrusion prevention inspection tool suitable for use in complex environments, characterized by: The small animal intrusion prevention inspection tool suitable for use in complex environments comprises an insulating operating rod and an industrial steering endoscope; a wire clamp is installed at one end of the insulating operating rod, and the other end is a handle end; the industrial steering endoscope comprises an endoscope host, and a camera connected to the endoscope host by wire; the wire clamp clamps the connection line between the endoscope host and the camera, and the clamping point is close to one end of the camera; the small animal intrusion prevention inspection tool suitable for use in complex environments moves the camera of the industrial steering endoscope into a narrow space through the insulating operating rod, and inspects small animal intrusion through the host of the industrial steering endoscope.

2. The small animal intrusion prevention inspection tool according to claim 1 is suitable for use in complex environments, characterized in that: The insulating operating rod comprises a plurality of rod sections, and each of the rod sections is connected via a steering positioning joint.

3. The small animal intrusion prevention inspection tool suitable for use in complex environments according to claim 1 is characterized in that: The endoscope host includes a display screen and a display screen mounting seat; an electric steering mechanism is installed at one end of the camera close to the connecting cable, and a lighting component is also provided inside the camera in addition to the camera component; the control switch of the electric steering mechanism is installed on the display screen mounting seat; the connecting cable is a Type-C data cable.

4. A method for inspecting small animals in a complex environment, the method for inspecting small animals in a complex environment is based on the anti-small animal intrusion inspection tool applicable to a complex environment as claimed in any one of claims 1 to 3, characterized in that: The following steps are involved: Step 1: Connect the equipment: Connect the industrial steering endoscope to the insulating operating rod, and connect it to the endoscope host with a Type-C data cable; turn on the endoscope host and check whether the equipment connection is normal; Step 2 Enter the inspection area: Carefully enter the inspection area according to the confirmed entry point; pay attention to ensure the safety of the surrounding environment and avoid potential electrical hazards; Step 3: Introduce the endoscope: Hold the insulating operating rod and slowly insert the connected endoscope into the inspection area, trying to avoid collision between the camera and surrounding objects; use the length and flexibility of the insulating operating rod to insert the camera into corners, gaps or pipes that are difficult to observe; Step 4: Image observation and adjustment: Observe the image transmitted by the camera in real time on the display screen of the endoscope host; use the operation buttons to adjust the brightness, contrast and magnification of the image to ensure clear observation; slowly move the insulating operating rod to adjust the angle and position of the endoscope, and comprehensively scan the inspection area, especially the channels and gaps where small animals may enter; Step 5 Record and analyze: Look for signs of small animal intrusion, such as hair, feces, and bite marks; If any suspicious signs are found, further zoom in on the image and observe in detail; use the host's recording function to record important scenes for subsequent analysis and archiving.

5. The method for detecting small animal intrusion in a complex environment according to claim 4, characterized in that: The endoscope host can perform artificial intelligence image recognition, which uses a deep learning algorithm to process and analyze the images collected by the endoscope in real time to realize small animal type recognition and intrusion situation determination; The specific steps are as follows: Step 501, data acquisition and preprocessing: the camera acquires real-time video stream, and the endoscope host receives image data; The image is preprocessed, including denoising, normalization, cropping, and enhancement to ensure that the image quality is suitable for model processing; Step 502, model training: constructing a classification model based on a convolutional neural network (CNN), and using a data set containing a variety of small animal categories and background noise for training; Step 503, real-time image analysis: the collected images are input into the trained model, features are extracted and classified; the recognition results are displayed in real time on the display screen of the endoscope host, and synchronously transmitted to the background host for storage and further analysis; Step 504, alarm triggering and response: judging whether there is any small animal intrusion behavior based on the model recognition result; if intrusion is detected, triggering sound alarm, visual alarm and background notification; Step 505, data update and optimization: by collecting data subsequently, the model is continuously updated to improve the recognition accuracy.

6. The method for detecting small animal intrusion in a complex environment according to claim 5, characterized in that: In step 501, when the endoscope host performs preprocessing on the image, the image is first grayed to reduce the computational complexity by graying; The calculation formula for grayscale is: I gray =0.299R+0.587G+0.114B In the formula, I gray Represents the pixel value of the grayscale image; R represents the pixel value of the red channel; G represents the pixel value of the green channel; B represents the pixel value of the blue channel; The endoscope host normalizes the image and maps the pixel values ​​to the interval [0, 1]; The normalized calculation formula is: In the formula, I norm Represents the normalized pixel value; I represents the pixel value to be normalized; Finally, the endoscope host crops and scales the image to unify the image size.

7. The method for detecting small animal intrusion in a complex environment according to claim 5, characterized in that: In step 502, the model structure of the convolutional neural network (CNN) classification model includes: a convolution layer for extracting low-level features, a multi-layer convolution and pooling layer for extracting high-level semantic features, and a fully connected layer for outputting classification results; Among them, the calculation formula of the feature extraction process is as follows: F=ReLu(W*I norm +b) Where F is the feature map; W is the weight matrix of the convolution kernel; b is the bias term corresponding to each convolution kernel; * is the convolution operation; I norm Represents the normalized pixel value; ReLu(x)=max(0,x): is the activation function; The calculation formula of the classification process is as follows: Where P(c) is the probability of category c; Z c is the output score of category c; exp is the mathematical exponential operation; It represents the sum of the output scores of all categories after being processed by the exponential function, n is the number of all possible categories, and i represents each category.

8. The method for detecting small animal intrusion in a complex environment according to claim 5, characterized in that: In step 504, if the probability of a certain type of small animal in the classification result P(c)>T; T is the set threshold, T=0.8, then an alarm is triggered; the alarm mode includes: the screen highlights the intrusion target location, sound prompts, and data synchronization to the background host.

9. The method for detecting small animal intrusion in a complex environment according to claim 5, characterized in that: In step 502, the model training process is as follows: Data preparation: Data set D = {(X i ,y i )} includes images of various small animals and their labels; D represents a set containing multiple samples; X i ,y i Represents the features and labels of the i-th sample in the dataset; performs data augmentation by rotation, flipping, and scaling to increase sample diversity; Model training: The loss function uses intersection and entropy loss: Where L represents the average cross entropy loss of the entire data set; m is the number of samples; n is the number of categories; y ij is the true category label of sample i; y ij is the probability predicted by the model; log is usually the natural logarithm; Optimize using the Adam optimizer: W,b←W,b-η·▽L Where W represents the weight matrix in the neural network or model; b is the bias term corresponding to each convolution kernel; η is the learning rate; ▽L represents the gradient of the loss function L with respect to the model parameters, including the weight W and the bias b; In step 503, the collected images are input into the trained model to extract features and classify them. Specifically, the endoscope host performs reasoning on each frame of the image by preprocessing the image; inputs the model and outputs the category probability; and finally determines whether to trigger an alarm based on the result.

Images