Intelligent camera cleaning system based on deep learning

Abstract

The invention relates to the technical field of intelligent camera cleaning, in particular to an intelligent camera cleaning system based on deep learning. The camera intelligent cleaning system based on deep learning comprises a camera flushing cleaning system and a lens smudginess judgment system. The camera flushing and cleaning system comprises a camera, a flushing device and an automatic control system, the camera is used for collecting images, the flushing device is used for spraying water to clean a lens, and the automatic control system is used for controlling starting and stopping of the flushing device and the water spraying amount; the lens smudginess judgment system adopts a deep learning method to judge whether the lens is smudginess or not. Compared with the prior art, the camera flushing and cleaning system has the advantages of being high in cleaning efficiency, stable in cleaning effect, high in adaptability, low in cost, high in intelligent degree and the like, and the requirements of the modern science and technology field for high-performance cameras can be better met.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent camera cleaning, and specifically, to an intelligent camera cleaning system based on deep learning. Background Art

[0002] In the field of modern technology, cameras have been widely used in various fields, such as security monitoring, intelligent transportation, medical imaging, industrial inspection, etc. With the development of technology, the performance of cameras has also been continuously improved, and there have been significant improvements in aspects such as resolution, frame rate, and sensitivity. However, the performance of a camera is closely related to the cleanliness of its lens. Once the lens is covered with dust, oil, or other contaminants, it will affect the image quality and reduce the performance of the camera. Therefore, keeping the lens of the camera clean is an important task.

[0003] Existing lens cleaning technologies mainly rely on physical cleaning methods, such as manual wiping, using air blowing for cleaning, ultrasonic cleaning, etc. Although these methods can effectively clean the lens, they all require manual intervention, have low efficiency, and are prone to causing secondary contamination to the lens. In addition, these methods cannot achieve real-time monitoring and automatic cleaning, and cannot meet the requirements in some special environments, such as high temperature, high pressure, corrosive gases, etc.

[0004] Therefore, the existing lens cleaning technologies have the following problems: First, the cleaning efficiency is low, manual intervention is required, and all-weather real-time monitoring and automatic cleaning cannot be achieved; second, the cleaning effect is unstable, and it is easy to cause secondary contamination to the lens; third, it cannot adapt to complex environments, such as high temperature, high pressure, corrosive gases, etc.; fourth, the cost is relatively high, and professional cleaning equipment and personnel are required. Therefore, the existing lens cleaning technologies can no longer meet the requirements of high-performance cameras in the field of modern technology. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent camera cleaning system based on deep learning to solve the problems of low cleaning efficiency, unstable cleaning effect, inability to adapt to complex environments, and relatively high cost mentioned in the above background art.

[0006] To achieve the above purpose, the present invention provides an intelligent camera cleaning system based on deep learning, including a camera flushing and cleaning system and a lens dirt determination system;

[0007] The camera flushing and cleaning system includes a camera, a flushing device, and an automatic control system. The camera is used to collect images, the flushing device is used to spray water to clean the lens, and the automatic control system is used to control the start and stop of the flushing device and the amount of water sprayed;

[0008] The lens dirt determination system uses deep learning methods to determine whether the lens is dirty.

[0009] As a further improvement of this technical solution, the lens dirt judgment system uses a deep learning algorithm to judge whether the lens is dirty, which specifically includes the following steps:

[0010] S1. Image acquisition;

[0011] S2. Image preprocessing;

[0012] S3. Feature extraction;

[0013] S4. Model training;

[0014] S5. Real-time prediction.

[0015] As a further improvement of this technical solution, the specific operation method for image acquisition in step S1 is: Deploy the camera in the area to be monitored, and collect images in real time through the camera. Images with and without dirt conditions should be collected. The resolution of the collected images is set to 1920x1080, and the frame rate is set to 30 frames per second.

[0016] As a further improvement of this technical solution, the specific operation method for image preprocessing in step S2 is: Preprocess the collected images, including denoising, grayscale conversion, and image enhancement. The size of the preprocessed images is adjusted to 256x256.

[0017] As a further improvement of this technical solution, the specific operation method for feature extraction in step S3 is: Use a deep learning model to extract the features of the preprocessed images. The features include edge features, texture features, and color features. The size of the extracted features is 32x32.

[0018] As a further improvement of this technical solution, the specific operation method for model training in step S4 is: Use the labeled training dataset to train the deep learning model. The training dataset includes clean lens images and dirty lens images. The deep learning model uses a convolutional neural network (CNN). The model structure includes 5 convolutional layers, 3 pooling layers, and 2 fully connected layers. During the model training process, use the cross-entropy loss function and the Adam optimizer. The learning rate is set to 0.001, and the number of training epochs is set to 100.

[0019] As a further improvement of this technical solution, the specific operation method for real-time prediction in step S5 is: Use the trained deep learning model to perform real-time prediction on the preprocessed images collected in real time to judge whether the lens is dirty. The prediction result includes the dirt probability. When the dirt probability is greater than the set threshold, it is judged that the lens is dirty.

[0020] As a further improvement of the technical solution, the flushing device uses a high-pressure nozzle, the water spraying pressure is set to 0.5 MPa, and the water spraying volume is set to 50 ml / time.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] 1. The camera flushing and cleaning system of the present invention can achieve all-weather real-time monitoring and automatic cleaning without manual intervention, greatly improving the cleaning efficiency. Especially in some special environments such as high temperature, high pressure, corrosive gases, etc. where manual cleaning is impossible, the automatic cleaning function of the present invention is particularly important.

[0023] 2. The cleaning system of the present invention uses a deep learning algorithm to judge whether the lens is dirty and sprays water for cleaning according to the judgment result, which can ensure more stable cleaning effect of the lens. At the same time, due to automatic cleaning, secondary pollution caused by manual cleaning can be avoided.

[0024] 3. The cleaning system of the present invention can work normally in complex environments without being affected by environmental conditions such as temperature, humidity, and air pressure. At the same time, due to automatic cleaning, damage caused by manual cleaning can be avoided.

[0025] 4. The cleaning system of the present invention adopts a simple mechanical structure and control system, with low cost, easy to promote and apply. At the same time, due to automatic cleaning, labor costs can be reduced and the overall operation cost can be lowered.

[0026] 5. The cleaning system of the present invention uses a deep learning algorithm for image recognition and judgment, with high intelligence level, can adapt to different lens pollution situations, and achieve more accurate cleaning. Description of the Drawings

[0027] Figure 1 It is the overall principle block diagram of the intelligent cleaning system for cameras based on deep learning of the present invention.

[0028] Figure 2 It is the operation flow chart of the intelligent cleaning system for cameras based on deep learning of the present invention. Detailed Embodiments

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0030] In a specific embodiment, as Figure 1As shown in the figure, the present invention provides an intelligent camera cleaning system based on deep learning, including a camera flushing cleaning system and a lens dirt judgment system.

[0031] Among them, the camera flushing cleaning system includes a camera, a flushing device, and an automatic control system. The camera is used to collect images, the flushing device is used to spray water to clean the lens, and the automatic control system is used to control the start and stop of the flushing device and the water spray volume.

[0032] The lens dirt judgment system uses deep learning methods to judge whether the lens is dirty. This method includes steps such as image acquisition, image preprocessing, feature extraction, model training, and prediction. First, images are collected through the camera, then the images are preprocessed, then the features of the images are extracted, and then the trained deep learning model is used for prediction to judge whether the lens is dirty.

[0033] In actual applications, first train the deep learning model required for the lens dirt judgment system:

[0034] Step 1: Image acquisition. Use the camera to be placed in the area to be monitored, and collect images in real time through the camera. Images with dirt and without dirt should be collected. The resolution of the collected images is set to 1920x1080, and the frame rate is set to 30 frames per second.

[0035] Step 2: Image preprocessing. Preprocess the collected images, including denoising, grayscale conversion, image enhancement, etc. The size of the preprocessed images is adjusted to 256x256 for subsequent feature extraction and model training.

[0036] Step 3: Feature extraction. Use the deep learning model to extract the features of the preprocessed images. The features include edge features, texture features, color features, etc. The size of the extracted features is 32x32.

[0037] Step 4: Model training. Use the labeled training data set to train the deep learning model. The training data set includes clean lens images and dirty lens images. The deep learning model uses a convolutional neural network (CNN), and the model structure includes 5 convolutional layers, 3 pooling layers, and 2 fully connected layers. During the model training process, a cross-entropy loss function and an Adam optimizer are used, the learning rate is set to 0.001, and the number of training epochs is set to 100.

[0038] After the model is trained, the entire system can run. The steps are as follows (see the flow chart in the appendix Figure 2 ).

[0039] Step Five: Prediction. Use the trained deep learning model to preprocess the images collected in real time and then perform real-time prediction to determine whether the lens is dirty. The prediction result includes the probability of dirt. When the probability of dirt is greater than the set threshold (e.g., 0.5), it is determined that the lens is dirty.

[0040] When the deep learning model determines that the lens is dirty, the automatic control system controls the flushing device to start and spray water to clean the lens. The flushing device uses a high-pressure nozzle, the water spraying pressure is set to 0.5 MPa, and the water spraying volume is set to 50 ml / time.

[0041] In summary, compared with the prior art, the camera flushing and cleaning system of the present invention has the advantages of high cleaning efficiency, stable cleaning effect, strong adaptability, low cost, high intelligence level, etc., and can better meet the requirements for high-performance cameras in the field of modern technology.

[0042] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A camera intelligent cleaning system based on deep learning, characterized in that: Including camera flushing cleaning system and lens dirtiness judgment system; The camera flushing cleaning system includes a camera, a flushing device and an automatic control system, wherein the camera is used to collect images, the flushing device is used to spray water to clean the lens, and the automatic control system is used to control the start and stop of the flushing device and the amount of water sprayed; The lens dirtiness judgment system uses a deep learning method to judge whether the lens is dirty.

2. The camera intelligent cleaning system based on deep learning according to claim 1, characterized in that: The lens dirtiness judgment system uses a deep learning method to judge whether the lens is dirty, and specifically includes the following steps: S1, image acquisition; S2, image preprocessing; S3, feature extraction; S4, model training; S5. Real-time prediction.

3. The camera intelligent cleaning system based on deep learning according to claim 2 is characterized in that: The specific operation method for image acquisition in step S1 is: use a camera to post in the area to be monitored, and collect images in real time through the camera, both dirty and clean conditions should be collected, and the collected image resolution is set to 1920x1080, and the frame rate is set to 30 frames / second.

4. The camera intelligent cleaning system based on deep learning according to claim 2, characterized in that: The specific operation method of performing image preprocessing in step S2 is: preprocessing the collected image, including denoising, grayscale conversion and image enhancement, and adjusting the size of the preprocessed image to 256x256.

5. The camera intelligent cleaning system based on deep learning according to claim 2 is characterized in that: The specific operation method for feature extraction in step S3 is: using a deep learning model to extract features of the preprocessed image, the features include edge features, texture features and color features, and the extracted feature size is 32x32.

6. The camera intelligent cleaning system based on deep learning according to claim 2, characterized in that: The specific operation method for model training in step S4 is: use a labeled training data set to train a deep learning model, the training data set includes clean lens images and dirty lens images, the deep learning model adopts a convolutional neural network (CNN), and the model structure includes 5 convolutional layers, 3 pooling layers and 2 fully connected layers. During the model training process, a cross entropy loss function and an Adam optimizer are used, the learning rate is set to 0.001, and the number of training rounds is set to 100.

7. The camera intelligent cleaning system based on deep learning according to claim 2, characterized in that: The specific operation method for performing real-time prediction in step S5 is: use the trained deep learning model to pre-process the real-time acquired image and then perform real-time prediction to determine whether the lens is dirty. The prediction result includes the probability of dirtiness. When the probability of dirtiness is greater than the set threshold, it is determined that the lens is dirty.

8. The camera intelligent cleaning system based on deep learning according to claim 1, characterized in that: The flushing device adopts a high-pressure nozzle, the water spraying pressure is set to 0.5MPa, and the water spraying volume is set to 50ml / time.

Images