Field oral and maxillofacial diagnosis and treatment vehicle with fresh air system

Abstract

According to the field oral and maxillofacial diagnosis and treatment vehicle with the fresh air system, the fresh air system is intelligently adjusted according to data collected by ultra-low power consumption information collection equipment such as a temperature sensor, an oxygen sensor and an environment visual sensor assembled on the vehicle, and the purposes of meeting medical requirements, saving energy and prolonging continuous operation time are achieved.

Description

Technical Field

[0001] The content of the present invention belongs to the field of medical devices. In particular, it relates to a field oral and maxillofacial diagnosis and treatment vehicle for use in high-altitude areas. Background Art

[0002] Mobile diagnosis and treatment vehicles are important facilities for ensuring the safety of personnel's lives in high-altitude areas, providing diagnosis and treatment environments such as oxygen therapy, heat preservation, and fresh air for personnel in urgent situations. Due to being located on the plateau with insufficient energy supply, mobile medical vehicles usually operate in a self-powered manner using solar energy, generators, etc. However, systems such as fresh air for maintaining the above functions have relatively high power, thus facing problems of energy shortage and electromagnetic exposure of high-power targets. And due to harsh natural conditions, the fresh air system must maintain a suitable temperature environment and oxygen environment in the diagnosis and treatment vehicle to ensure the safety of the entire medical process, especially the surgical process. How to control the fresh air system with low power consumption and accurately provide a suitable environment for the diagnosis and treatment vehicle is an urgent problem to be solved.

[0003] In the prior art, multiple sensors are used for fusion to control the fresh air system, but most of these sensors involve physical sensors such as temperature, humidity, and air pressure, and the collected signals are all environmental signals. However, due to the characteristics of small space and dense personnel in the diagnosis and treatment vehicle, if the influencing factors of personnel are ignored, the control of the fresh air system cannot be accurately achieved. The prior art also proposes to achieve precise control through neural networks, but these neural networks often have complex models and require high computing power, which further causes an increase in power consumption. Summary of the Invention

[0004] By designing an intelligent system with the function of collecting and perceiving information about the surrounding environment, through information collection and perception, intelligent control is implemented on the fresh air system of the mobile medical vehicle, achieving the goals of meeting medical needs, saving energy, extending the continuous operation time, reducing power consumption, and improving environmental adaptability.

[0005] A field oral and maxillofacial diagnosis and treatment vehicle with a fresh air system, including a vehicle body and a diagnosis and treatment compartment. The diagnosis and treatment compartment is equipped with an air fresh air system, a central control system, and an information collection system; Among them, the central control system conducts intelligent control on the air fresh air system. The specific method is as follows: (1) The heating (cooling) capacity is adjusted according to the following formula: , where represents the heating amount (positive value) or cooling amount (negative value) output by the controlled fresh air system per unit time, represents the desired constant temperature to be achieved, represents the currently measured temperature, is a linear proportional coefficient, represents the influence of each person on the change in environmental temperature, is the number of people in the current environment, represents the adjustment coefficient; (2) The oxygen production amount is adjusted according to the following formula: , where represents the oxygen production amount per unit time for controlling the output of the fresh air system, represents the expected oxygen concentration to be achieved, represents the currently measured oxygen concentration, is the proportional coefficient, is the personal oxygen consumption constant, is the number of people in the vehicle; The central controller takes the complete image inside the vehicle compartment and the image of the door part as input data and inputs them into the neural network model, where represents the spatial coordinates of a pixel of the image, represents the number of frames; The neural network model includes 5 hidden layers and 1 output layer; The first hidden layer is: Among them, , are the linear mapping coefficients of the first hidden layer, , are the corresponding linear biases, is the feature of the mapping space column of the first hidden layer ( ), is the activation function; The output layer is: , where , represents the behavior classification category, is the natural exponential function. is the linear mapping coefficient, is the corresponding linear bias, is the output of the fifth hidden layer; According to the above behavior recognition result, the current number of people in the vehicle is calculated in real time .

[0006] , where is the adjustment coefficient reference, represents the percentage of the remaining battery power, is the natural logarithm function.

[0007] The second hidden layer is: Among them, , are the linear mapping coefficients of the second hidden layer, , is the corresponding linear offset, is the feature of the second hidden layer mapping space row ( ).

[0008] The third hidden layer is: Among them, , are the linear mapping coefficients of the third hidden layer, , is the corresponding linear offset, is the feature of the third hidden layer mapping time ( ).

[0009] The fourth hidden layer is: Among them, , are independent linear convolution kernels respectively, is the linear offset of the fourth hidden layer, The value range of 1 - 16 indicates 16 independent convolution kernels in each group.

[0010] The fifth hidden layer is: are the linear mapping coefficients of the fifth hidden layer, is the corresponding linear offset, and the output of the fifth hidden layer is a 64 - dimensional vector, that is The value range is 1 - 64.

[0011] Inside the medical treatment carriage, there are also: a treatment processing table, diagnostic equipment, and treatment equipment.

[0012] The information acquisition system includes a temperature sensor, an oxygen sensor, and an environmental vision sensor.

[0013] The environmental vision sensor is an infrared - band camera.

[0014] The medical treatment vehicle is applied to high - altitude areas.

[0015] The invention points and technical effects of the present invention: 1. The invention point of the present invention lies in innovatively proposing that the influence of personnel on the temperature and oxygen in the narrow space of the medical treatment vehicle cannot be ignored. Therefore, it is necessary to intelligently control the fresh - air system according to the number of personnel, provide more accurate temperature and oxygen, reduce the power consumption of the fresh - air system, save energy, and thereby reduce the electromagnetic target exposure risk of the whole vehicle.

[0016] 2. The invention point of the present invention also lies in optimizing the specific control methods of temperature control and oxygen control, enabling the medical treatment vehicle to adapt to the rescue of the wounded under harsh natural conditions in high - altitude areas and providing a suitable rescue environment.

[0017] 3. The inventive point of the present invention also lies in optimizing the recognition model for the number of people in the medical treatment vehicle. Through the design of the model structure, loss function, etc., and using two images of the entire vehicle compartment and the door as inputs, real-time and accurate statistics of the number of people are achieved while consuming low power. And by introducing the above-mentioned behavior data, constraints are increased, and the uncertainties brought by environmental noise, overlapping of people, etc. are reduced, so that a relatively simple model can also achieve good recognition effects. Detailed implementation manners

[0018] A field oral and maxillofacial medical treatment vehicle with a fresh air system includes a vehicle body and a medical treatment compartment. The vehicle body is used to realize vehicle driving, vehicle control, and carry the medical treatment compartment. The interior of the medical treatment compartment includes: a treatment processing table, diagnostic equipment, treatment equipment, a sewage treatment system, an air fresh air system, a central control system, and an information collection system.

[0019] Among them, the information collection system is an ultra-low power consumption information collection device, including a temperature sensor, an oxygen sensor, and an environmental vision sensor. The above sensors collect information such as the temperature inside the vehicle, the oxygen content inside the vehicle, and environmental images, and after sampling and quantization, form digital temperature data, oxygen content data, environmental image data, etc., and send them to the central control system as the data source for its intelligent control of the fresh air system.

[0020] The above temperature data, oxygen content data, and environmental image data are collected regularly at a certain frequency. Among them, the temperature data and oxygen content data are directly input into the control system. After the environmental image data is further processed, characteristic quantities reflecting oxygen consumption and temperature influence are formed and input into the control system. The control system makes intelligent control based on the above inputs.

[0021] Particularly, the environmental image data is an infrared band image, and the environmental vision sensor is an infrared band camera.

[0022] In order to achieve the energy-saving goal, the above sensors and the control system are both designed with ultra-low power consumption.

[0023] The central control system makes intelligent control of the air fresh air system to achieve low-power and high-precision fresh air control. The specific control method is as follows: 1. Based on the temperature data, oxygen content data, and environmental image data, intelligently calculate and control the oxygen generation and heating of the fresh air system, save energy consumption, and improve comfort.

[0024] 2. The temperature data and oxygen content data are used as the current basis, and the environmental image data is used as the future inference basis. Compared with the general fresh air adjustment control system, it not only has the function of balancing the current environmental parameters, but also has the ability to adaptively adjust according to environmental changes.

[0025] The above method is characterized in that by collecting environmental image data, perceiving the number of people in the environment, and then obtaining the heat generation and oxygen consumption of people, the fresh air system is adaptively adjusted according to the above environmental impact variables.

[0026] The adjustable control quantities of the fresh air system include (1) heating (cooling) capacity and (2) oxygen generation capacity.

[0027] (1) The heating (cooling) capacity is adjusted according to the following formula The left side of the formula represents the heating capacity (positive value) or cooling capacity (negative value) per unit time of the output of the fresh air system. The right side of the equation represents the desired constant temperature, represents the currently measured temperature, obtained by the temperature sensor. is the linear proportional coefficient, indicating the energy consumption (heat / cooling) required to change the current environmental temperature by one degree, obtained by prior experimental calibration, and is a constant. represents the influence of each person on the change of environmental temperature, obtained by prior experimental calibration, and is a constant. This constant is calibrated separately according to different seasons and different environmental characteristics (such as temperature, humidity) and formulated into a table. The constant is obtained by looking up the table according to the on-site environment during actual use. Therefore, the corresponding constant is selected according to the current environment where the medical treatment vehicle is located. is the number of people in the current environment, obtained by calculation by the visual analysis system according to the data collected by the environmental vision sensor according to the analysis model (detailed below). represents the adjustment coefficient. The larger the adjustment coefficient, the higher the heating (cooling) capacity per unit time, and the higher the energy consumption of the fresh air system. The adjustment coefficient is calculated according to the remaining power (percentage) of the power supply of the mobile medical vehicle.

[0028] is the adjustment coefficient reference, obtained by prior experimental calibration, and is a constant. represents the percentage of the remaining battery power. The battery is full at 1 and exhausted at 0, is the natural logarithm function.

[0029] (2) The oxygen generation capacity is adjusted according to the following formula.

[0030] The left side of the formula represents the oxygen generation capacity per unit time of the output of the fresh air system. The right side of the equation represents the desired oxygen concentration, represents the currently measured oxygen concentration, measured by the oxygen sensor. is a proportionality coefficient, which is directly proportional to the size of the space inside the medical vehicle and inversely proportional to the oxygen demand rate, and is a constant that can be obtained through pre-experiment testing. is an individual oxygen consumption constant, which is obtained through pre-experiment calibration. This constant is also calibrated according to different working environments (such as altitude, air pressure, etc.) and compiled into a table, and this constant can be obtained by looking up the table according to the on-site environment during actual use. The number of people in the vehicle is the same as above.

[0031] Of course, the above method is a preferred implementation mode adopted by the present invention. In fact, if the accurate number of people in the carriage can be obtained, according to the calibrated influence degree of personnel on temperature and oxygen, other calculation methods can also be used to obtain the control formula of the fresh air system, and it is not necessarily required to follow the above formula. One of the first contributions of the present invention is to propose the influence of the number of people on temperature and oxygen control.

[0032] In a complex event environment (such as rescuing the wounded), the personnel in the diagnosis and treatment carriage often change. Due to the narrow space inside the carriage, different from the open space, the number of people will not only affect the oxygen consumption inside the carriage, but also affect the temperature inside the carriage. Therefore, the fast, accurate and intelligent statistics of the number of people in the carriage has an important impact on accurately controlling the fresh air system of the air. For this reason, the present invention proposes the following method to realize the real-time statistics of the personnel situation in the vehicle, so that the intelligent fresh air system can adaptively adjust the heating (cooling) capacity and oxygen production according to the number of people.

[0033] Use an infrared camera to capture the interior environment of the vehicle, and use infrared image analysis methods to count the number of people in the vehicle.

[0034] Person target recognition is a classic task in the field of machine vision. Currently, the main methods include template- and correlation-based recognition methods, and recognition methods based on machine learning models. The former method is relatively simple, with fewer model parameters and high computational efficiency, but has a relatively large recognition error. The latter model is relatively complex, with more parameters and cumbersome calculations, but has a lower recognition error. In recent years, recognition methods based on neural network models have made great progress. However, in order to achieve better recognition performance, relatively complex models are required, and the hidden layers usually reach hundreds of layers. The mobile medical vehicle is located on the plateau, with limited communication and energy equipment, and it is impossible to transmit image data back to the server side for in-depth analysis. Moreover, the ultra-low-power central control system installed on the vehicle is also unable to run deep models with a large number of parameters. Excessive improvement of the processing power of the central processing unit will lead to increased energy consumption, making it difficult to adapt to harsh environments such as plateaus and mountainous areas, and it is more likely to be undesirably discovered and tracked due to heat source and electromagnetic factors. Similarly, a large amount of image communication will also undesirably expose the medical vehicle. In particular, the space in the medical vehicle is narrow and the personnel are relatively crowded, and there is a certain degree of occlusion in image acquisition, which poses a great challenge to accurate recognition. Therefore, an innovative visual analysis model is proposed. According to the environmental characteristics of the medical vehicle, it has been specially modified and optimized compared with the existing model, greatly streamlining the network structure and the number of parameters, and enabling the personnel target recognition task to be completed online in the ultra-low-power system of the medical vehicle.

[0035] Use an infrared camera to take images inside the vehicle, and the camera shooting area completely covers each seat inside the vehicle. The complete image obtained is denoted as .

[0036] Specifically, install another camera to take pictures of the door area, or the aforementioned camera can capture the door area. The image of the door area is denoted as .

[0037] Existing recognition methods usually take an image to recognize the number of people in it, and are easily interfered by environmental noise, personnel occlusion, etc. The present invention obtains images from two angles, including not only the information of the people inside the vehicle, but also particularly the image change information at the door. The behavior of people getting on and off the vehicle is used as part of the model inference. By introducing the above-mentioned behavior data, the constraints are increased, and the uncertainty brought by environmental noise, personnel occlusion, etc. is reduced, so that a relatively streamlined model can also achieve good recognition results.

[0038] As described above, the input data of the model includes the complete image and the image of the door area.

[0039] Among them, represents the spatial coordinates of a pixel in the image, Indicates the number of frames, that is, each set of input data consists of several consecutive images to model and identify the changes of personnel.

[0040] The calculation method of the model hidden layer is as follows.

[0041] The first hidden layer: Among them, and are the linear mapping coefficients of the first hidden layer, and are the corresponding linear biases. The value ranges are both 1 - 32. The first hidden layer maps the features of the column ( ), and unifies the column dimensions of the two types of images to 32. is the activation function: is the proportional control constant, preferably . Equation 6 can improve the robustness of personnel recognition in images and the recognition accuracy compared with classical activation functions such as Sigmoid and ReLU, and is more suitable for the small space of the carriage and the situation of overlapping personnel.

[0042] The second hidden layer: Among them, and are the linear mapping coefficients of the second hidden layer, and are the corresponding linear biases. The value ranges are both 1 - 32. The second hidden layer maps the features of the row ( ), and unifies the row dimensions of the two types of images to 32. is the activation function (as above).

[0043] The third hidden layer: Among them, and are the linear mapping coefficients of the third hidden layer, and are the corresponding linear biases. The value ranges are both 1 - 16. The third hidden layer maps the features of the time ( ). is the activation function (as above).

[0044] Through the first, second, and third hidden layers, the features of the three independent dimensions of the input data are mapped to a finite space of a set of constant dimensions, so as to capture features through the convolutional network and identify the behavior of people entering and leaving.

[0045] Fourth hidden layer: Among them, 、 are independent linear convolutional kernels respectively, is the linear bias of the fourth hidden layer, The value of 1-16 indicates 16 independent convolutional kernels in each group, and the size of the convolutional kernel is 7*7*5. Each group corresponds to the features from the complete image and the features of the car door image respectively. is the activation function (as above).

[0046] Fifth hidden layer: is the linear mapping coefficient of the fifth hidden layer, is the corresponding linear bias. The output of the fifth hidden layer is a 64-dimensional vector, that is The value range is 1-64. is the activation function (as above).

[0047] The letters in the above brackets are all the independent variables of the corresponding functions.

[0048] The model output layer is as follows.

[0049] Among them represents the behavior classification category, and each dimension output corresponds to getting on the car, getting off the car, and remaining unchanged respectively. is the natural exponential function. is the linear mapping coefficient, is the corresponding linear bias, is the activation function.

[0050] According to the cost function is the class label of the training sample, the corresponding class is 1, and the rest are 0. After training, the above model is used to identify the input data, and the dimension corresponding to the maximum output value is taken as the behavior recognition category (getting on the car, getting off the car, remaining unchanged).

[0051] According to the above behavior recognition results, the number of people in the vehicle is counted in real time, and the current number is returned to formulas (1) and (3) The present invention provides a fresh air intelligent control system for a mobile medical vehicle in high altitude areas. According to the data collected by ultra-low power consumption information collection devices assembled on the medical vehicle, such as temperature sensors, oxygen sensors, environmental vision sensors, etc., the fresh air system is intelligently adjusted to achieve the goals of meeting medical needs, saving energy, and extending the continuous operation time.

[0052] Table 1 Table 2

Claims

1. A field oral and maxillofacial diagnosis and treatment vehicle with a fresh air system, characterized by: It includes a vehicle body and a diagnosis and treatment compartment, wherein the diagnosis and treatment compartment is equipped with an air fresh air system, a central control system, and an information collection system; The central control system intelligently controls the fresh air system. The specific method is as follows: (1) The heating (cooling) amount is adjusted according to the following formula: ,in, Indicates the heating capacity (positive value) or cooling capacity (negative value) per unit time output by the fresh air system. Indicates the desired constant temperature. Indicates the currently measured temperature. is the linear scaling factor, Indicates the impact of each person on the change of ambient temperature, is the number of people in the current environment, Indicates the adjustment coefficient; (2) The oxygen production is adjusted according to the following formula: ,in Indicates the oxygen production per unit time output by the fresh air system. Indicates the desired oxygen concentration. Indicates the currently measured oxygen concentration. is the proportionality coefficient, is the personal oxygen consumption constant, is the number of people in the car; The central controller collects the complete image of the interior of the car collected by the information collection system and door area images As input data into the neural network model, Represents the spatial coordinates of a pixel in the image, Indicates the number of frames; the neural network model includes 5 hidden layers and 1 output layer; the first hidden layer is: in, , is the linear mapping coefficient of the first hidden layer, , is the corresponding linear bias, Mapping spatial columns for the first hidden layer ( )feature, is the activation function; The output layer is: ,in , Indicates the behavior classification category, is a natural exponential function. are the linear mapping coefficients, is the corresponding linear bias, is the output of the fifth hidden layer; According to the above behavior recognition results, the current number of people in the car is calculated in real time .

2. The diagnostic vehicle according to claim 1, characterized in that: ,in is the adjustment coefficient benchmark, Indicates the remaining battery power percentage. is the natural logarithm function.

3. The diagnostic vehicle according to claim 1, characterized in that: The second hidden layer is: in, , is the linear mapping coefficient of the second hidden layer, , is the corresponding linear bias, Mapping space rows for the second hidden layer ( )feature.

4. The diagnostic vehicle according to claim 3, characterized in that: The third hidden layer is: in, , is the linear mapping coefficient of the third hidden layer, , is the corresponding linear bias, Mapping time for the third hidden layer ( )feature.

5. The diagnostic vehicle according to claim 4, characterized in that: The fourth hidden layer is: in, , are independent linear convolution kernels, is the linear bias of the fourth hidden layer, The value 1-16 indicates 16 independent convolution kernels in each group.

6. The diagnostic vehicle according to claim 5, characterized in that: The fifth hidden layer is: is the linear mapping coefficient of the fifth hidden layer, is the corresponding linear bias, and the output of the fifth hidden layer is a 64-dimensional vector, that is, The value range is 1-64.

7. The diagnostic vehicle according to claim 1, characterized in that: The interior of the diagnosis and treatment carriage also includes: a treatment table, diagnostic equipment, and treatment equipment.

8. The diagnostic vehicle according to claim 1, characterized in that: The information collection system includes temperature sensors, oxygen sensors, and environmental vision sensors.

9. The diagnostic vehicle according to claim 8, characterized in that: The environmental visual sensor is an infrared band camera.

10. The diagnostic vehicle according to claim 1, characterized in that: The diagnosis and treatment vehicle is used in high-altitude areas.

Images