Auto-orientation of image acquisition equipment for containment appearance based on center mark recognition

Abstract In this paper, we present an image acquisition device for building surface. To position the image and generate an unfolded view of the building, the camera needs to be orientated to determine the camera’s pose before image acquisition. In this paper, we propose an automatic orientation method based on target point recognition. Traditional orientation methods often have low success rates and poor accuracy. Our method divides the orientation task into three steps: detecting the prism and its centre point sequentially, and finally completing the orientation. To address the challenge of detecting prism center points (PCPs), we add the scale sequence feature fusion and ScaleConcat modules based on YOLOv8s, achieving multi-scale feature fusion to enrich semantic and positional information. Additionally, adding a small object detection head enhances the detection accuracy for small targets such as PCPs. Experimental results demonstrate that our method achieves a mean average precision (mAP) of 94.68% for prism detection, while achieving precision and recall of 89.12% and 88.40% respectively for PCP detection, with a mAP of 92.68%. Furthermore, our method exhibits high orientation accuracy, with horizontal and vertical angle detection errors of just <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:msup> <mml:mn>4.61</mml:mn> <mml:mrow> <mml:mi>′</mml:mi> <mml:mi>′</mml:mi> </mml:mrow> </mml:msup> </mml:mrow> </mml:math> and <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:msup> <mml:mn>2.93</mml:mn> <mml:mrow> <mml:mi>′</mml:mi> <mml:mi>′</mml:mi> </mml:mrow> </mml:msup> </mml:mrow> </mml:math> respectively. The proposed model offers fast orientation speed, high success rates, and precise orientation, meeting the requirements of practical applications.