Cotton gin fault detection method based on image recognition

Abstract

The invention belongs to the technical field of cotton gin safety monitoring, and particularly discloses a cotton gin fault detection method based on image recognition, which comprises the following steps: acquiring an initial cotton roll image of an opening area of a cotton roll cavity in real time by using a visual camera, performing noise reduction on the initial cotton roll image information to obtain a cotton roll outer surface image, and calculating the cotton gin fault according to the cotton roll outer surface image. The method comprises the following steps: acquiring a plurality of groups of lap flow velocity values, judging whether the cotton gin is in a cotton jamming fault state or not according to the plurality of groups of lap flow velocity values, if so, performing cotton jamming degree identification by utilizing a preset convolutional neural network model, determining a cotton jamming degree grade, determining a corresponding fault early warning operation from a preset early warning strategy, and performing early warning according to the fault early warning operation. The cotton clamping state is determined through double fault analysis of the cotton roll flow speed and the cotton roll image, different fault early warning operations are correspondingly set, automatic identification and control are achieved in the whole process, the accuracy and timeliness of cotton clamping identification are improved, the cotton ginning efficiency is improved, and operation safety is guaranteed.

Description

Technical Field

[0001] The invention relates to the technical field of safety monitoring of cotton gins, and in particular to a cotton gin fault detection method based on image recognition. Background Art

[0002] Cotton gin is a kind of cotton processing machinery, which is a machine for separating lint from cotton seed. Commonly used cotton gins include sawtooth cotton gin and roller cotton gin. Among them, the sawtooth cotton gin is a machine that uses a high-speed rotating (circumferential speed of about 12-13 m / s) disc saw blade to hook and pull cotton fibers through the gap between ribs to separate them from cotton seeds. There are two types of brush type and airflow type. Its main working parts include feeding rollers, cleaning mechanism, cotton ginning work box, cotton ginning ribs, saw blade cylinder, brush roller or airflow suction nozzle, cotton collecting box, etc. The cotton seed is fed into the cleaning mechanism through the paired feeding rollers, and then enters the front box of the cotton gin. It is thrown to the saw blade by the cotton picking roller, and large debris such as bell shells are leaked. The cotton seed is hooked by the saw teeth and brought into the cotton ginning work box. The saw blade rotates rapidly, driving the mutually pulled cotton seed to form a rotating cotton seed roll. The saw teeth hook the fiber and rotate. After passing through the gap between adjacent ginning ribs, the cotton fiber is brushed off by the brush roller (or airflow) and sent to the cotton collecting box; the cotton seeds are blocked by the ribs, move down along the rib surface between the two saw blades, and are discharged out of the machine through the cotton seed comb.

[0003] During the operation of the cotton gin, cotton jams are prone to occur at the junction of the saw blade cylinder and the gin ribs. Once cotton jams occur, the cotton will heat up and catch fire in a very short time, causing equipment failure. Currently, cotton jams are identified through manual observation. Due to the limited recognition of the human eye and the long reaction time of the human body, it is impossible to identify in time and perform relevant preventive operations, resulting in frequent cotton jams and fires in the cotton gin, which reduces the cotton ginning efficiency and poses a safety hazard. Summary of the invention

[0004] The object of the present invention is to provide a cotton gin fault detection method based on image recognition to solve the problem that cotton jams in the cotton gin cannot be identified in time.

[0005] To achieve the above object, the basic solution provided by the present invention is: a cotton ginning machine fault detection method based on image recognition, comprising:

[0006] S1: using a visual camera to obtain an initial cotton roll image of the cotton roll cavity opening area in real time, and reducing noise of the initial cotton roll image information to obtain an image of the outer surface of the cotton roll;

[0007] S2, obtaining multiple groups of cotton roll flow rate values ​​within a preset time period using a visual camera based on the cotton roll outer surface image;

[0008] S3: judging whether the cotton gin is in a cotton jam fault state according to the multiple sets of cotton roll flow rate values; if in a cotton jam fault state, identifying the fault degree based on the cotton roll outer surface image using a preset convolutional neural network model to determine the cotton jam degree level of the current cotton gin; wherein the cotton jam degree level includes a primary cotton jam fault and a secondary cotton jam fault;

[0009] S4: According to the cotton jam level, the corresponding fault warning operation is determined from the preset warning strategy;

[0010] S5: Issue a warning according to the fault warning operation.

[0011] Since cotton is easily jammed at the junction of the serrated cylinder and the ginning ribs, the rotation speed of the cotton roll will slow down and the outer surface of the cotton roll will bulge or collapse in the cotton jam state. In addition, since the junction of the serrated cylinder and the ginning ribs is inside the cotton roll cavity and cannot be directly detected, the present invention uses a visual camera to monitor the cotton roll image in the opening area of ​​the cotton roll cavity in real time, and uses the functional coding inside the visual camera to obtain multiple groups of cotton roll flow rate values ​​within a preset time period. According to the multiple groups of cotton roll flow rate values, it is determined whether the cotton gin is in a cotton jam fault state. If it is in the cotton jam fault state, the preset convolutional neural network model is used to identify the degree of fault, determine the cotton jam degree level of the current cotton gin, and interlock the cotton jam degree level with the fault warning operation. Different fault states correspond to different fault warning operations, which improves the practicality of fault identification. The whole process is automatically identified and controlled, which solves the problem that cotton jams in the cotton gin cannot be identified in time, improves cotton ginning efficiency, and ensures the safe operation of the cotton gin.

[0012] Preferably, the denoising of the initial cotton roll image in step S1 specifically includes:

[0013] Performing wavelet transform on the initial cotton roll image to obtain the high frequency part and the low frequency part of the initial cotton roll image;

[0014] The part of the high-frequency part where the wavelet coefficient is lower than the preset wavelet coefficient threshold is set to zero to obtain the high-frequency part after noise reduction;

[0015] The low-frequency part and the high-frequency part after noise reduction are reconstructed by wavelet to obtain the outer surface image of the cotton roll.

[0016] It is understandable that useful signals are manifested as low-frequency parts or stable signals, while noise is manifested as high-frequency signals. The cotton image is transformed by wavelet transform to remove unnecessary interference information in the high-frequency part, realize denoising of the initial cotton roll image, and improve the clarity and reliability of the image.

[0017] Preferably, the determination of the cotton gin failure in step S3 specifically includes:

[0018] Arrange multiple groups of cotton roll flow rate values ​​in chronological order;

[0019] Comparing the arranged multiple groups of cotton roll flow rate values ​​with the preset fault speed interval to determine whether the multiple groups of cotton roll flow rate values ​​continuously fall within the preset fault speed interval;

[0020] If it does not continuously fall within the preset fault speed range, it is determined that the current cotton gin is not stuck; if it continuously falls within the preset fault speed range, it is determined that the current cotton gin is in a cotton stuck fault state.

[0021] By setting a preset fault speed interval, the obtained multiple groups of cotton roll flow rate values ​​are compared with the preset fault speed interval. If they are continuously within the interval, it means that there may be cotton jamming at the junction of the serrated cylinder and the cotton gin ribs. If they fall continuously within the preset fault speed interval, it is determined that the current cotton gin is in a cotton jamming fault state, and the degree of cotton jam is identified, providing a data basis for subsequent fault warning operations interlocked with the cotton jam degree level.

[0022] Preferably, the pre-training of the convolutional neural network model in step S3 specifically includes:

[0023] Obtain historical cotton jam images of the cotton roll cavity opening area;

[0024] Classify the historical cotton jam images to obtain primary cotton jam images and secondary cotton jam images;

[0025] Based on the primary cotton jam images and the secondary cotton jam images, a training set and a validation set are established;

[0026] The convolutional neural network model is trained using the training set to obtain a trained convolutional neural network model;

[0027] The validation set is used to test whether the trained convolutional neural network model is qualified. If qualified, the convolutional neural network model is output; if unqualified, the training set is used again to train the convolutional neural network model until the convolutional neural network model is qualified.

[0028] Based on historical cotton jam images, the image information is divided into primary cotton jam images and secondary cotton jam images. The primary cotton jam corresponds to a low cotton jam degree, so it is set to primary cotton jam. The secondary cotton jam corresponds to a high cotton jam degree, so it is set to secondary cotton jam. Based on these two types of images, the training set and validation set of the model are established. The training set is used to train the initial convolutional neural network model, and then the validation set is used to verify whether the trained convolutional neural network model is qualified, so as to obtain a model that can realize image fault recognition and improve the accuracy of model fault recognition.

[0029] Preferably, the preset convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer;

[0030] The process of using the preset convolutional neural network model to identify the fault degree in step S3 specifically includes the following steps:

[0031] Normalizing the outer surface image of the cotton roll to obtain standard image information; wherein the standard image information includes pixel point information and color channel information;

[0032] Input the standard image information into the input layer;

[0033] The convolution layer receives the standard image information output by the input layer, and uses a preset convolution kernel and ReLU function as its activation function to extract features from the standard image information to obtain the initial features of the image;

[0034] The pooling layer receives the initial image features output by the convolution layer and performs average pooling on the initial image features to obtain the target image features;

[0035] The fully connected layer receives the target image features output by the pooling layer, flattens the target image features into a one-dimensional vector, and uses the Softmax function to classify them to obtain the probability corresponding to each cotton jam level category, and selects the cotton jam level category corresponding to the maximum probability, and uses the cotton jam level category corresponding to the maximum probability as the cotton jam level corresponding to the current cotton gin image;

[0036] The output layer receives the cotton jam level output by the fully connected layer and outputs the cotton jam level.

[0037] The inner layers of the convolutional neural network model are sequentially set to the input layer, the convolution layer, the pooling layer, the fully connected layer and the output layer. Since the convolutional neural network model uses the gradient descent algorithm for learning, the input features need to be standardized. Therefore, before the cotton roll outer surface image is input into the input layer, the present invention performs normalization on the cotton roll outer surface image to meet the data processing conditions of the convolutional neural network model, and then inputs the standard image information into the input layer. The input layer transmits the data to the convolution layer, and uses a convolution kernel of a preset size and a ReLU function as its activation function to extract image features. Then, the pooling layer performs average pooling processing to obtain target image features, and then inputs them into the fully connected layer to flatten the target image features into a one-dimensional vector, and uses the Softmax function for classification to obtain the cotton jam degree level corresponding to the current cotton gin image. Then, the output layer outputs the cotton jam degree level to realize cotton jam fault analysis based on the cotton roll outer surface image, avoid the limitations and hysteresis of current manual recognition, and improve the efficiency and accuracy of cotton jam fault recognition.

[0038] Preferably, the fault warning operation corresponding to the first-level cotton jam fault in the preset warning strategy specifically includes:

[0039] Generate and send a first-level signal light flashing instruction to the signal light, and generate corresponding first-level voice broadcast content, and use the broadcasting equipment to broadcast according to the first-level voice broadcast content to achieve first-level signal light warning and first-level voice broadcast warning.

[0040] In case of a first-level cotton jam, the cotton rolling speed may slow down, and a slight bulge or depression may appear on the outer surface of the cotton roll, but the cotton gin can still operate normally. The degree of the fault is low, so the present invention only performs voice broadcast warning and signal light warning to remind the operator that the cotton gin may have a slight cotton jam.

[0041] Preferably, the fault warning operation corresponding to the secondary cotton jam fault in the preset warning strategy specifically includes:

[0042] Generate and send a cotton gin unpacking instruction to the cotton gin, so that the cotton gin unpacks;

[0043] Generate and send a secondary signal light flashing command to the signal light, and generate corresponding secondary voice broadcast content, and use the broadcasting device to broadcast according to the secondary voice broadcast content to realize secondary signal light prompts and secondary voice broadcast prompts.

[0044] In case of secondary cotton jam, a larger cotton ball may be stuck between the serrated cylinder and the cotton ginning ribs, resulting in a significant reduction in the cotton rolling speed and a large bulge or depression on the outer surface of the cotton roll. The present invention directly controls the cotton gin to open the box and uses signal lights and voice broadcast prompts to prompt the operating workers to deal with the cotton jam problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a diagram of the internal structure of a cotton gin. DETAILED DESCRIPTION

[0046] The present invention is further described in detail below through specific embodiments:

[0047] The reference numerals in the drawings of the specification include: 1. cotton gin; 2. serrated cylinder; 3. cotton gin ribs; 4. visual camera; 5. cotton roll cavity.

[0048] like Figure 1 As shown, this embodiment provides a cotton ginning machine fault detection method based on image recognition, which specifically includes the following contents:

[0049] S1: using the visual camera 4 to obtain the initial cotton roll image of the opening area of ​​the cotton roll cavity 5 in real time, and reducing the noise of the initial cotton roll image information to obtain the cotton roll outer surface image.

[0050] Since cotton is easily jammed at the junction of the serrated cylinder 2 and the ginning ribs 3, the rotation speed of the cotton will slow down in the cotton jam state, and since the junction of the serrated cylinder 2 and the ginning ribs 3 is inside the cotton roll cavity 5 and cannot be directly detected, in this embodiment, the visual camera 4 is loaded on the outer shell wall of the cotton gin 1, and the camera is aimed at the opening area of ​​the cotton roll cavity 5 for shooting, and the visual camera 4 captures the image through the camera.

[0051] The denoising of the initial cotton roll image information specifically includes: performing wavelet transform on the initial cotton roll image to obtain the high-frequency part and the low-frequency part of the initial cotton roll image, setting the part of the high-frequency part whose wavelet coefficient is lower than the preset wavelet coefficient threshold to zero to obtain the denoised high-frequency part, performing wavelet reconstruction on the low-frequency part and the denoised high-frequency part to obtain the cotton roll outer surface image, removing unnecessary interference information in the high-frequency part, realizing denoising of the cotton roll outer surface image, and improving the clarity and reliability of the image.

[0052] S2. Based on the outer surface image of the cotton roll, a visual camera is used to obtain multiple groups of cotton roll flow rate values ​​within a preset time period.

[0053] After the visual camera 4 acquires the image of the outer surface of the cotton roll, a mapping relationship between the image pixel coordinates and the world coordinates of the corresponding points in space is established by calibration, image features are extracted and selected, the actual displacement of the image features over a period of time is obtained, and then the cotton roll flow rate value is calculated.

[0054] S2: judging whether the cotton gin 1 is in a cotton jam fault state according to multiple groups of cotton roll flow rate values; if in a cotton jam fault state, identifying the degree of the fault based on the outer surface image of the cotton roll using a preset convolutional neural network model, and determining the cotton jam degree level of the current cotton gin 1; wherein the cotton jam degree level includes a first-level cotton jam fault and a second-level cotton jam fault.

[0055] In this embodiment, the degree of cotton jam is divided into primary cotton jam and secondary cotton jam. In the case of primary cotton jam, the flow rate on the outer surface of the cotton roll may slow down, and the outer surface of the cotton roll may slightly bulge or dent, but the cotton gin 1 can still operate normally, and the degree of failure is low. The cotton gin 1 can be left unpacked and continue to operate; in the case of secondary cotton jam, a large cotton ball may be stuck between the serrated cylinder 2 and the gin rib 3, resulting in a significant decrease in the flow rate on the outer surface of the cotton roll, and a large bulge or dent on the outer surface of the cotton roll. The degree of failure is high, and the cotton gin 1 needs to be unpacked to deal with the cotton jam. It should be noted that in this embodiment, the cotton gin 1 is controlled to be in a non-stop state when the cotton gin 1 is unpacked.

[0056] The identification of the cotton jam fault state specifically includes: arranging multiple groups of cotton roll flow rate values ​​in chronological order, and by setting a preset fault speed interval, comparing the obtained multiple groups of cotton roll flow rate values ​​with the preset fault speed interval. If it is within the interval, it means that there may be cotton jam at the junction of the serrated cylinder 2 and the cotton gin rib 3. If it falls continuously within the preset fault speed interval, it is determined that the current cotton gin is in a cotton jam fault state, thereby realizing the identification of the degree of cotton jam, so as to provide a data basis for subsequent fault warning operations interlocked with the cotton jam degree level.

[0057] The pre-training of the convolutional neural network model specifically includes: obtaining historical cotton jam images of the cotton roll cavity opening area, classifying the historical cotton jam images to obtain first-level cotton jam images and second-level cotton jam images, the first-level cotton jam corresponding to the outer surface of the cotton roll has a small bulge or depression, that is, the cotton jam degree is low, so it is set to a first-level cotton jam, and the second-level cotton jam corresponding to the outer surface of the cotton roll has a large bulge or depression, that is, the cotton jam degree is high, and it is set to a second-level cotton jam, and then based on the first-level cotton jam image and the second-level cotton jam image, a training set and a verification set are established, and the training set is used to train the convolutional neural network model to obtain the trained convolutional neural network model, and the verification set is used to check whether the trained convolutional neural network model is qualified. If qualified, the convolutional neural network model is output; if unqualified, the training set is used again to train the convolutional neural network model until the convolutional neural network model is qualified, so as to obtain a model that can realize image fault recognition and improve the accuracy of model fault recognition.

[0058] In this embodiment, the layers in the convolutional neural network model are sequentially set to the input layer, convolution layer, pooling layer, fully connected layer and output layer. Since the convolutional neural network model uses the gradient descent algorithm for learning, the input features need to be standardized. Therefore, in this embodiment, before the cotton roll outer surface image is input to the input layer, the cotton roll outer surface image is normalized to meet the data processing conditions of the convolutional neural network model. The normalized standard image information includes pixel information and color channel information.

[0059] The process of convolutional neural network fault recognition specifically includes: inputting the normalized standard image information into the input layer, the input layer transmits the data to the convolution layer, uses the preset size of the convolution kernel and the ReLU function as its activation function to extract image features, and then performs average pooling processing through the pooling layer to obtain the target image features, and then inputs the target image features into the fully connected layer to flatten the target image features into a one-dimensional vector, uses the Softmax function for classification, obtains the probability corresponding to each cotton jam degree category, screens out the cotton jam degree category corresponding to the maximum probability, and uses the cotton jam degree category corresponding to the maximum probability as the cotton jam degree level corresponding to the current cotton gin image, and then uses the output layer to output the cotton jam degree level to realize cotton jam degree analysis based on the outer surface image of the cotton roll, avoids the limitations and lags of current manual recognition, and improves the efficiency and accuracy of cotton jam fault recognition. For example, in this embodiment, the layers of the convolutional neural network model are set to input layer, convolution layer, pooling layer, convolution layer, pooling layer...convolution layer, pooling layer, full connection point and output layer, and a 32×32 standard image is input to the input layer. The convolution layer uses a 5×5 convolution kernel to convolve the image of the input layer, and then obtains 20 28×28 two-dimensional feature maps after the activation function ReLU. The pooling layer performs average pooling on all 2×2 sub-blocks in the convolution layer to obtain 20 14×14 sub-blocks. The following convolutional layer and pooling layer operations repeat the above steps; finally, 50 5×5 2D feature maps are obtained. The fully connected layer flattens the target image features into a one-dimensional vector, and uses the Softmax function for classification to obtain the probability corresponding to each cotton jam level category, and selects the cotton jam level category with the maximum probability, and uses the cotton jam level category with the maximum probability as the cotton jam level corresponding to the current cotton gin image, and then uses the output layer to output the cotton jam level. Among them, the excitation function ReLU and the Softmax function are shown in Formula 1 and Formula 2 respectively:

[0060] f(x)=max(0,x) (1)

[0061] Among them, x is the input, and the output is a larger value between x and 0, which effectively solves the problem of gradient disappearance.

[0062]

[0063] Among them, z j represents the score of the j-th jam level category, softmax(z) j It represents the estimated probability of the j-th cotton jamming degree category. K is the vector dimension output by the softmax function. The sum of the vector elements is 1. Finally, the probability values ​​corresponding to each cotton jamming degree level are output.

[0064] S3: According to the cotton jam level, the corresponding fault warning operation is determined from the preset warning strategy.

[0065] The fault warning operation corresponding to the first-level cotton jam fault in the preset warning strategy is: generate and send a first-level signal light flashing instruction to the signal light, and generate a corresponding first-level voice broadcast content, and use the broadcasting equipment to broadcast according to the first-level voice broadcast content to achieve a first-level signal light warning and a first-level voice broadcast warning to remind the operating worker that the cotton gin 1 may have a slight cotton jam.

[0066] The fault warning operation corresponding to the second-level cotton jam fault in the preset warning strategy is: generate and send the cotton gin 1 unpacking instruction to the cotton gin 1, so that the cotton gin 1 is unpacked, and then further generate and send the second-level signal light flashing instruction to the signal light, and generate the corresponding second-level voice broadcast content, and use the broadcasting equipment to broadcast according to the second-level voice broadcast content, to realize the second-level signal light prompt and the second-level voice broadcast prompt, to prompt the operating worker to deal with the cotton jam problem.

[0067] S5: Issue a warning according to the fault warning operation.

[0068] The above is only an embodiment of the present invention, and the common knowledge such as the known specific structure and characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several deformations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A cotton ginning machine fault detection method based on image recognition, characterized in that: include: S1: using a visual camera (4) to obtain an initial cotton roll image of the opening area of ​​the cotton roll cavity (5) in real time, and reducing noise on the initial cotton roll image information to obtain an image of the outer surface of the cotton roll; S2, obtaining multiple groups of cotton roll flow rate values ​​within a preset time period using the visual camera (4) based on the cotton roll outer surface image; S3: judging whether the cotton gin (1) is in a cotton jam fault state according to the plurality of groups of cotton roll flow rate values; if in a cotton jam fault state, identifying the degree of the fault based on the cotton roll outer surface image using a preset convolutional neural network model, and determining the cotton jam degree level of the current cotton gin (1); wherein the cotton jam degree level includes a primary cotton jam fault and a secondary cotton jam fault; S4: Determine a corresponding fault warning operation from a preset warning strategy according to the cotton jam level; S5: Produce a warning according to the fault warning operation.

2. A cotton ginning machine fault detection method based on image recognition according to claim 1, characterized in that: In step S1, the initial cotton roll image is subjected to noise reduction, which specifically includes: Performing wavelet transform on the initial cotton roll image to obtain a high-frequency part and a low-frequency part of the initial cotton roll image; The part of the high-frequency part whose wavelet coefficients are lower than a preset wavelet coefficient threshold is set to zero to obtain a high-frequency part after noise reduction; The low-frequency part and the high-frequency part after noise reduction are subjected to wavelet reconstruction to obtain an image of the outer surface of the cotton roll.

3. A cotton ginning machine fault detection method based on image recognition according to claim 1, characterized in that: The determination of the cotton gin failure in step S3 specifically includes: Arranging the plurality of groups of cotton roll flow rate values ​​in chronological order; Comparing the arranged multiple groups of cotton roll flow rate values ​​with the preset fault speed interval to determine whether the multiple groups of cotton roll flow rate values ​​continuously fall within the preset fault speed interval; If it does not continuously fall within the preset fault speed range, it is determined that the current cotton gin (1) is not stuck; if it continuously falls within the preset fault speed range, it is determined that the current cotton gin (1) is in a cotton stuck fault state.

4. A cotton ginning machine fault detection method based on image recognition according to claim 1, characterized in that: The pre-training of the convolutional neural network model in step S3 specifically includes: Acquire a historical cotton jam image of the cotton roll cavity (5) opening area; Classify the historical cotton jam images to obtain primary cotton jam images and secondary cotton jam images; Based on the primary cotton jam image and the secondary cotton jam image, establishing a training set and a validation set; Using the training set to train the convolutional neural network model to obtain a trained convolutional neural network model; The validation set is used to check whether the trained convolutional neural network model is qualified. If qualified, the convolutional neural network model is output; if unqualified, the training set is re-used to train the convolutional neural network model until the convolutional neural network model is qualified.

5. A cotton ginning machine fault detection method based on image recognition according to claim 4, characterized in that: The preset convolutional neural network model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; The process of using the preset convolutional neural network model to identify the fault degree in step S3 specifically includes the following steps: Normalizing the cotton roll outer surface image to obtain standard image information; wherein the standard image information includes pixel point information and color channel information; Inputting the standard image information into the input layer; The convolution layer receives the standard image information output by the input layer, and uses a convolution kernel of a preset size and a ReLU function as its activation function to extract features from the standard image information to obtain initial features of the image; The pooling layer receives the initial image features output by the convolution layer, and performs average pooling processing on the initial image features to obtain target image features; The fully connected layer receives the target image features output by the pooling layer, flattens the target image features into a one-dimensional vector, classifies them using the Softmax function, obtains the probability corresponding to each cotton jam level category, selects the cotton jam level category corresponding to the maximum probability, and uses the cotton jam level category corresponding to the maximum probability as the cotton jam level corresponding to the current cotton gin image; The output layer receives the cotton jam level output by the fully connected layer, and outputs the cotton jam level.

6. A cotton ginning machine fault detection method based on image recognition according to claim 1, characterized in that: The fault warning operations corresponding to the first-level cotton jam fault in the preset warning strategy specifically include: Generate and send a first-level signal light flashing instruction to the signal light, and generate corresponding first-level voice broadcast content, and use the broadcasting device to broadcast according to the first-level voice broadcast content to achieve first-level signal light warning and first-level voice broadcast warning.

7. A cotton ginning machine fault detection method based on image recognition according to claim 1, characterized in that ,The fault warning operations corresponding to the secondary cotton jam fault in the preset warning strategy specifically include: Generate and send a cotton gin unpacking instruction to the cotton gin (1), so that the cotton gin (1) unpacks; Generate and send a secondary signal light flashing instruction to the signal light, and generate corresponding secondary voice broadcast content, and use the broadcast device to broadcast according to the secondary voice broadcast content to realize the secondary signal light prompt and the secondary voice broadcast prompt.

Images