基于计算机视觉的停车位车辆存在性检测方法

为了提高停车场的车位利用率,有效地管理停车场,提出了1种基于计算机视觉的车位检测方法.与传统的视频检测方法不同,该方法在停车位上绘制了特定的辅助识别图案,图案具有各向同质性的特征,在大部分光照、阴影的影响下具备图案特征不变性,且与一般车辆上绘制的图案有显著差别,在此基础上采用图像检测算法对图案进行数学解析描述,作为检测目标,同时采用图像识别算法,逐行扫描各个像素,利用模式匹配判断停车位状态.通过选取停车场的2个车位进行实例验证,准确率为98.12%.结果表明该方法识别速度快、准确率高,具有很好的应用前景. 

Parking Space Occupancy Detection Based on Computer Vision

In order to improve the utilization of parking spaces and manage parking spaces more effectively, a detec- tion technique of existing parking vehicles based on computer vision is proposed. Compared with traditional video detec- tion techniques, this method draws specific homogeneous graphics, which are significantly different from ordinary graph- ics painted on vehicles and remain constant under the influence of light and shadow, on parking spaces to assist identifica- tion. Then, it mathematically describes the pattern by image detection algorithm, and sets the patterns as the targets for pixel-by-pixel scan in line by image recognition algorithm, and determines the status of parking spaces through pattern matching. Finally, two parking spaces are selected to test the computer-vision-based detection method, and its accuracy is 98.12%. The result shows that the method can recognize the target fast with high accuracy, which has a good prospect.