Source: Elecfans  19^th Aug 2025

The field of intelligent driving is currently experiencing a strategic competition over the routes of perception technology, with lidar and pure vision solutions representing two distinct technological paths. Lidar provides precise distance and spatial information through high-precision 3D point clouds, while pure vision relies on algorithms to process image data to simulate human visual perception. Each has its own advantages and disadvantages, and both have their own advantages in different application scenarios. With technological progress and cost reduction, the industry is gradually forming a stratified technical route: entry-level models adopt pure vision solutions to control costs, while high-end models enhance safety through multi-sensor fusion. This trend not only reflects the objective laws of technological evolution but also demonstrates the market's precise grasp of the demands of consumers in different price segments. In the future, the integration of pure vision and lidar will become the mainstream direction, seeking the best balance between cost and safety through the collaborative optimization of software and hardware. Meanwhile, the improvement of policies and regulations will provide clearer guidance for the selection of technical routes. 

Differences in technical principles and perception methods

the point cloud processing algorithms of LiDAR and the deep learning algorithms of vision solutions are constantly being optimized. However, there are still significant differences between the two in terms of data types, processing methods, and information expression. 

Comparison of Cost, Reliability and Environmental Adaptability

In terms of cost, there is a significant gap between lidar and pure vision solutions. Early lidar was expensive, with a single unit costing over 10,000 yuan, while the cost of cameras in pure vision solutions was only a few dozen dollars, giving pure vision solutions a clear advantage in hardware cost. However, with the maturation and mass production of solid-state lidar technology, this gap is narrowing. By 2025, the mass production price of mainstream solid-state lidar (such as Q*** S3) has dropped to the range of 100 to 1,000 US dollars, and some manufacturers (such as Shenzhen Li*) aim for a target price of 6,000 to 8,000 yuan per unit, making the application of lidar in mid-to-high-end vehicles more feasible.

In terms of reliability, LiDAR can eliminate point cloud distortion caused by bumps through IMU (Inertial Measurement Unit) fusion technology, thereby enhancing positioning accuracy. For instance, the tightly coupled positioning method based on MSCKF that integrates IMU and LiDAR can effectively address the issue of point cloud distortion and achieve high-precision mapping and positioning in outdoor scenarios. In contrast, pure vision solutions optimize complex scene recognition through end-to-end deep learning models to reduce false triggers. The AI-driven intelligent driving system of the car model launched in 2025 has optimized computing power, raising the total computing power to 508 TOPS, which is 4 to 6 times that of the mainstream level. Coupled with large models on the vehicle end and ultra-high-definition environmental perception technology, it provides a safer and more continuous intelligent driving experience. 

Strategies for Technology Route Selection and Market Promotion Dynamics of Automobile Manufacturers

In 2025, domestic automakers' choices of intelligent driving perception technology routes showed a clear stratified strategy. Xiaoyu Automobile adopted a pure vision solution on entry-level models, eliminating lidar to control costs, with the starting price dropping to 176,800 yuan; while on high-end models, lidar was still retained, forming a stratified strategy of "pure vision for entry-level and multi-sensor for high-end". 

Huawei adopts a more refined market stratification strategy. Its ADS 4.0 system still comes standard with lidar (4 units) in high-end models, building a 360° all-round environmental perception system; while mid-range models use a vision-based solution, forming a market layout of "high-end fusion, mid-range pure vision". Through the integration of lidar, vision perception, millimeter-wave and ultrasonic technologies, Huawei has constructed multi-dimensional perception capabilities. Its GOD (General Obstacle Detection) network and PDP (Prediction, Decision-making and Control) network perform well in complex scenarios. For instance, in a 110-kilometer test, there were only three takeovers, including having to cross double yellow lines to detour around road construction, failing to recognize the changing traffic light, and misjudging a straight-ahead red light when turning left. 

The strategy of Baidu Apollo's rapid development is also worth analyzing. Despite rumors that Baidu will shift to a pure visual approach, its fully driverless test in Shanghai in 2025 still relies on a multi-sensor fusion solution, such as the installation of lidar on the Extreme Edition. However, in the project in Dubai, Baidu explicitly uses four Hesa lidars. 

Other automakers stick to the LiDAR route. ET is equipped with a 1550nm hybrid solid-state LiDAR with a detection range of up to 500 meters, highlighting the advantage of long-distance perception. Traditional automakers like BYD and BAIC, however, adopt a more cautious fusion approach, gradually advancing technological upgrades while ensuring safety. Even Aion has launched a "triple LiDAR solution", becoming the only model in the 300,000-yuan class to use three radars, emphasizing the concept of "safety for all". 

Trends of Technology Convergence and Future Development Directions 

The future development directions are mainly reflected in four aspects: solid-state, miniaturization, cost reduction and intelligence. Solid-state lidar improves reliability and reduces costs by reducing mechanical components; the trend of miniaturization enables lidar to be better integrated into vehicle body design; cost reduction is achieved through large-scale production and technological innovation; intelligence is reflected in the optimization of algorithms and the improvement of data fusion capabilities. 

The Impact of Policies and Regulations on the Development of Intelligent Driving 

Policies and regulations not only influence the selection of technical routes but also promote the development of vehicle-road coordination, pilot projects in closed scenarios, and the construction of national-level testing bases. For instance, the Ministry of Industry and Information Technology has officially initiated the international standard project "evaluation of Road Vehicle Autonomous Driving System Test Scenarios and Generation of Test Cases" (ISO 34505), which will fill the gap in international standards for generating test cases in the test scenarios of autonomous driving systems. 

In addition, data security regulations also have an impact on the technological routes of intelligent driving. In May 2021, the Cyberspace Administration of China released the "Several Provisions on the Security Management of Automotive Data (Draft for Comment)", which defined the data generated by intelligent connected vehicles, clarified issues such as the responsible party, data scope, collection methods, privacy protection, and data export. This has prompted automakers to pay more attention to data security and privacy protection when choosing their technological routes. For instance, Huawei has adopted a fully self-developed solution to achieve data closed-loop and improve optimization efficiency. 

Market penetration rate and industry development trends

According to market research, the market size of automotive lidar in China is expected to reach 24 billion yuan by 2025, with a penetration rate exceeding 30%. This is mainly attributed to the maturity and mass production of solid-state lidar technology, as well as the significant reduction in costs. 

The market for intelligent driving sensors is taking on a diversified pattern. According to predictions, the global market size for sensor modules in intelligent driving vehicles will reach 36 billion US dollars by 2030. Among them, ultrasonic sensors, 360-degree panoramic cameras and front cameras will remain the mainstream, with expected market sizes of 12 billion US dollars, 8.7 billion US dollars and 6.9 billion US dollars respectively; the market size for radars will reach 12.9 billion US dollars, including 7.9 billion US dollars for long-range radars and 5 billion US dollars for short-range radars. 

Conclusion

Lidar and pure vision solutions are not mutually exclusive but rather complementary and coexisting technical routes. Lidar provides high-precision three-dimensional spatial information, while pure vision captures rich semantic information. The integration of the two can build a more comprehensive and reliable environmental perception system. With the maturation of solid-state lidar technology and the reduction in costs, as well as the continuous optimization of pure vision algorithms, the fusion solution will become the mainstream direction of intelligent driving, especially in L3 and above autonomous driving scenarios. 

In the future, with technological advancements and cost reductions, intelligent driving perception technology will evolve towards high performance, low cost, miniaturization and solid-state. At the same time, technologies such as vehicle-road coordination and V2X will also be deeply integrated with perception systems to build a safer, more efficient and intelligent transportation ecosystem. In this process, the fusion of lidar and pure vision solutions will play a key role, laying a solid foundation for the sustainable development of intelligent driving technology.