A bus door opening and closing state detection model training method, a detection method, a device and mobile terminal equipment

Abstract

The invention discloses a bus door opening and closing state detection model training method, a detection method, a device and mobile terminal equipment, and the method comprises the following steps: 1, collecting the position of a bus door, and RGB color images and depth images which are synchronous in time and position, and forming an image combination; 2, fusing the RGB color image and the depth image, using the fused image as the input of a classification neural network, and training the neural network until convergence; and step 3, adding a deep prediction branch structure to the classification neural network to form a neural network B, taking the RGB color image as the input of the neural network B, calculating the loss of deep prediction, calculating the loss of vehicle door state prediction, and training the neural network B until the loss converges. By fusing and using TOF depth image and RGB color image data, the convergence speed and generalization ability of the model are improved. And the neural network model embedded with the image depth information is obtained through training, so that the accuracy of bus door state prediction is effectively improved.

Description

technical field [0001] The invention relates to the fields of computer vision and intelligent public transportation, in particular to a method for detecting the opening and closing state of a bus door, and in particular to a model training method, a detection method, a device, and a mobile terminal device for detecting the opening and closing state of a bus door. Background technique [0002] Effective and accurate passenger flow statistics are of great significance to bus intelligent scheduling, intelligent scheduling, and line network optimization. The accuracy of passenger flow statistics often depends on the accuracy of bus door opening and closing detection. The passenger flow of getting on and off the bus belongs to the effective passenger flow, and the accurate realization of the station passenger flow statistics function is closely related to the effective detection of the bus door opening and closing status. Existing detection of the opening and closing state of the...

AI Technical Summary

Problems solved by technology

However, RGB color images are easily disturbed by external light. For example, when the light at the door is relatively strong, it is easy to misjudge the closed state of the car door as the open state of the car door.

Images