The Algorithm Research Of Mine Transport Vehicle Loading State Recognition Based On Image Processing

With the development of transportation,the problem of determining the loading state of mine transport vehicles has been paid more attention.At present,the loading status of mine transport vehicles completely depends on manual experience and intuitive judgment.The poor working environment and low working efficiency pose a great threat to the personal safety of field workers.In order to improve the efficiency of transport vehicles and realize the automatic identification of transport vehicles’ loading state,this paper presents an image processing based loading status recognition algorithm for mine transport vehicles based on the existing video surveillance.In order to make full use of the original monitoring equipment and reduce unnecessary waste,and considering that the camera installation environment is relatively simple in the mine environment,this paper proposes a monocular image-based algorithm research on the loading state of the mine transport vehicle.Firstly,the license plate recognition technology is used to identify the transport vehicles and eliminate the interference brought by the vehicles outside the mining area.Each license plate number corresponds to unique vehicle information.We can obtain the specific information of the vehicles in this mining area by the vehicle license plate number.The full convolutional networks(FCN)and Edgebox are used to detect the license plate position from rough to fine.The CNN structure is simplified by adopting the structural design of 2 convolutional layers and 2 pooling layers,and the global average pooling layer is used to replace the full connection layer,which improving the accuracy of license plate recognition and detection accuracy.Secondly,the 3D-Box of the transport vehicle is obtained by using the three-dimensional reconstruction technology based on monocular image,which provides a basis for the subsequent angle transformation and the identification of loading state.The improvement is made on the basis of the Faster-RCNN algorithm.For example,FPN is added as a feature extraction subnet,and a rectangular convolution kernel is used to make it more suitable for the shape characteristics of transport vehicles.Added direction prediction branch,and obtained a more accurate 2D-Box bounding box.According to the fusion of the size information and surface features of the vehicle,and the 2D-Box of the vehicle as the constraint condition,then the 3D-Box of the transport vehicle can be obtained.Finally,introducing perspective transformation algorithm,the four vertices on the side of the 3D-Box is taken as the origin which can reconstruct the image from the monitoring perspective into a side view.The picture from the monitoring perspective is reconstructed into a side view.The loading state is determined according to the load rate of goods and the coordinate position relation between the highest loading point of goods and the front plate assembly of the transport vehicle.The experimental results show that the method proposed in this paper is superior to other methods in that it is less affected by the external environment and has high detection accuracy.It can be carried out on the original equipment,effectively reducing the cost of investment,avoid manual on-site operation,reduce the risk of personal safety and has high practical value.